Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

High Carbon Stock forests provide co‐benefits for tropical biodiversity

High Carbon Stock forests provide co‐benefits for tropical biodiversity INTRODUCTIONAgricultural expansion has emerged as a pervasive threat to tropical forests and biodiversity (Wilcove, Giam, Edwards, Fisher, & Koh, ), and has been implicated in the loss of c. 150 million ha of tropical forest over the last three decades (Gibbs et al., ; Hansen et al., ). A key driver of recent deforestation has been rising demand for cheap vegetable oil such as that from oil palm (Elaeis guineensis), which now covers 16 million ha across 43 countries, often at the expense of tropical forest (Pirker, Mosnier, Kraxner, Havlík, & Obersteiner, ).The potential economic and social benefits associated with oil palm (Potter, ) contrast with severe and well‐documented ecological impacts. Conversion of forest to oil palm plantation results in major biodiversity decline, which disproportionately affects forest specialists and species of conservation concern, resulting in assemblages dominated by disturbance‐tolerant generalists (Fitzherbert et al., ; Yaap, Struebig, Paoli, & Koh, ). With around 19% of land suitable for oil palm coinciding with areas of high biodiversity (Pirker et al., ), across forested Asia, Africa and South America, the full ecological impact of this commodity crop is yet to be fully realised. Mitigation measures that reconcile environmental sustainability, biodiversity conservation and production of crops such as oil palm are therefore essential in tropical regions.Retaining native habitat in oil palm estates is known to enhance the biological value of plantation landscapes by providing ecological refugia and improved connectivity (Gillies & St Clair, ; Struebig et al., ). However, in practice, the designation of conservation set‐asides can be hindered by agricultural profitability, with income exceeding US$11,240/ha over a 25 year growing cycle (Fisher, Edwards, Giam, & Wilcove, ). Thus, conservation efforts seeking to preserve forest within plantations may be more successful when economic incentives are provided to offset the opportunity costs associated with foregoing development. Amongst several mitigation tools available, two incentive‐driven policies based on carbon stocks have gained traction in tropical regions: (1) REDD+ (United Nations Reducing Emissions from Deforestation and forest Degradation) and related carbon credit schemes, and (2) improved land‐use planning via commodity certification (Yaap et al., ).REDD+, a payment for ecosystem services tool to mitigate climate change, aims to compensate stakeholders in developing nations for conservation initiatives and sustainable management practices that protect and restore the carbon sequestered by forests (Venter & Koh, ). If REDD+ were to achieve its economic potential, payments generated could make forest conservation financially competitive compared to oil palm cultivation (Butler, Koh, & Ghazoul, ). REDD+ is also attractive to conservation because it may deliver co‐benefits, whereby safeguarding high carbon areas also protects biodiversity at no additional cost (Gardner et al., ). However, this assumes spatial congruence between areas of high carbon and biodiversity. In reality, it is difficult to generalise on the nature, strength and extent of these co‐benefits because outcomes vary both within and between spatial scales (e.g. global: Naidoo et al., vs. Strassburg et al., ; national: Egoh, Reyers, Rouget, Bode, & Richardson, vs. Murray, Grenyer, Wunder, Raes, & Jones, ; landscape: Ruiz‐Jaen & Potvin, vs. Kessler et al., ). The extent to which carbon‐biodiversity co‐benefit assumptions hold at management unit scales appropriate to decision‐makers remains an open question.REDD+ is largely implemented at sub‐national levels. While an increasing number of studies are recognising the importance of fine‐scale assessments (e.g. Beaudrot et al., ; Magnago et al., ; Sollmann et al., ), most information on biodiversity co‐benefits is derived from global‐ and national‐scale studies that demonstrate overreliance on coarse‐grained, secondary data sources. Carbon data are typically derived from global maps (e.g. Avitabile et al., ; Baccini et al., ), which have limited application at local‐scales pertinent to management (Mitchard et al., ). Furthermore, field‐based species data are widely underrepresented in the co‐benefits literature due to the costs associated with biodiversity surveys in the tropics (Gardner et al., ). Researchers predominantly rely on coarse species range delineations, which are fraught with uncertainty (Rodríguez‐Castañeda, Hof, Jansson, & Harding, ) and may not account for localised extirpation due to anthropogenic pressure (Harrison et al., ). Despite statistical advances that account for imperfect detection in biodiversity indices (Royle & Dorazio, ), these methods have received relatively limited application in a co‐benefits context (but see Gilroy et al., ; Sollmann et al., ), resulting in possible underestimates of species assemblages. Consequently, biodiversity co‐benefits assessments at local‐scales, using primary, fine‐grained data would provide valuable policy insights.While the potential importance of REDD+ cannot be overstated, agricultural certification schemes show promise to ensure sustainable practices as companies benefit from greater access to environmentally conscious markets and increased price premiums of certified products (Yaap et al., ). The Roundtable on Sustainable Palm Oil (RSPO) is often seen as an exemplar scheme within the agricultural sector, currently certifying 21% of the global palm oil market across 2.48 million ha of land (RSPO, ). RSPO certification prohibits the conversion of high conservation value habitat in oil palm estates. However, associated assessment procedures have attracted criticism, raising concerns that current methodologies do not afford adequate biodiversity protection (Edwards, Fisher, & Wilcove, ; Yaap et al., ).The High Carbon Stock (HCS) approach has emerged as a land‐use planning tool to demarcate conservation priority areas based on carbon value, and is being explored within the RSPO architecture and that of other certification schemes. The HCS methodology seeks to conserve biodiverse and ecologically functional forest networks within agricultural concessions by directing conversion towards heavily degraded land of low carbon value (Rosoman, Sheun, Opal, Anderson, & Trapshah, ). This is achieved by stratifying land into discrete classes according to vegetation density and structure, which are then adopted as proxies for above‐ground carbon stocks and assumed to support varying levels of biodiversity. These strata are subsequently validated using field‐derived above‐ground carbon estimates, before land parcels are prioritised for conversion based on area and connectivity (Rosoman et al., ).The HCS approach has attracted widespread interest amongst agricultural industries with 10 million ha of land being evaluated across five oil palm producing countries (G. Rosoman, unpubl. data). As a model scheme, the successful integration of the HCS Approach within the RSPO framework may encourage uptake across other certifiable tropical commodities, such as rubber and soya. Nevertheless, the extent to which HCS strata correspond to areas of high biodiversity value is dependent on the accurate partitioning of vegetation classes according to their carbon value, as well as the underlying association between carbon and biodiversity. Before the HCS approach is formally adopted within certification standards, these assumptions should be tested to understand the conservation merit of the tool.Here, we determine the effectiveness of incentive‐driven carbon‐based mechanisms to safeguard biodiversity, and provide the first validation of both the carbon and biodiversity credentials of the HCS land‐use planning tool. Our appraisal focuses on a landscape undergoing conversion from forest to oil palm in Borneo, a region characterised by high deforestation and forest degradation (Gaveau et al., ; Struebig et al., ) that is typical of most HCS applications. First, we validate the accuracy of the HCS classification procedure and quantify the biodiversity value of the vegetation strata. We then assess the potential for REDD+ to deliver biodiversity co‐benefits using primary and high‐resolution data sources. To assess the influence of spatial grain on the nature of co‐benefit relationships, we compare global‐ and local‐scale measures of carbon. Throughout, we employ biodiversity indices that explicitly account for imperfect detection to provide a more accurate representation of species assemblages than simple species counts. Our work evaluates the extent to which policy options that attach greater economic significance to conservation protect vulnerable tropical forests and safeguard biodiversity.MATERIALS AND METHODSStudy systemThe study was conducted over a 13,153 ha development area comprising the Stability of Altered Forest Ecosystems project (SAFE; www.safeproject.net) and surrounding plantations in Kalabakan Forest Reserve, Sabah, Malaysian Borneo (4°46′N, 116°57′ E; Figure ). SAFE is a landscape‐scale forest modification experiment (Ewers et al., ) comprising highly disturbed lowland and hill dipterocarp forest that was logged multiple times between 1978 and 2008. The wider landscape includes near‐pristine forest in Brantian‐Tatulit Virgin Jungle Reserve, twice‐logged forest in Ulu Segama Forest Reserve, and plantations (primarily oil palm).High Carbon Stock (HCS) classification of the study landscape in Sabah, Borneo. Forest cover was delineated into four strata on the basis of vegetation density (Dense Forest, Young Regenerating Forest, Scrub, Open Land) and supplemented with two reference classes (Continuous Logged Forest, Oil Palm) to act as forest and agricultural controls. Points indicate camera‐trap locations (N = 115) [Colour figure can be viewed at wileyonlinelibrary.com]HCS classification and validationThe HCS Approach uses high‐resolution remotely‐sensed images to stratify concessions into six vegetation classes, each with unique structural characteristics (in descending order of carbon value): (1) “High Density Forest”; (2) “Medium Density Forest”; (3) “Low Density Forest”; (4) “Young Regenerating Forest”; (5) “Scrub”; and (6) “Open Land” (see Appendix S1). In practice, the High, Medium and Low Density Forest strata are aggregated as “Dense Forest” and earmarked for conservation. Young Regenerating Forest can also comprise valuable carbon stocks and is also spared from development. The threshold for allocating land for production rests on distinguishing these strata from heavily degraded Scrub and Open Land. Therefore, we mapped Dense Forest, Young Regenerating Forest, Scrub and Open Land as separate classes.All spatial data processing was implemented in ArcGIS 10.2.1 (ESRI). We used Landsat 8 and SPOT5 satellite imagery (15 and 2.5 m resolution respectively; temporal range: 2012–2014) to stratify forest habitat using HCS assessment protocols (see Appendix S1). Multiple data sources were chosen to minimise classification difficulties associated with cloud cover and haze. We undertook a supervised classification of satellite images, supplemented with visual interpretation techniques to correct for the potentially confounding effects of topographic shadow (Wulder, Franklin, White, Cranny, & Dechka, ). The resulting classes were then calibrated using above‐ground carbon values derived from forest inventory data (N = 139), collected as part of the core SAFE monitoring programme. These data conform to standardised forest inventory protocols (http://www.rainfor.org), calculating carbon as a function of above‐ground biomass (trees >10 cm DBH) using an established pantropical algorithm (Chave et al., ). Resulting HCS classes were validated using independently derived carbon estimates (Pfeifer et al., ; see Appendix S2).Camera‐trap sampling of medium‐large mammalsWe delineated terrestrial mammal diversity as these taxa are consistently prioritised in policy, land‐use planning and certification schemes. Remotely operated digital cameras (HC500 Hyperfire, Reconyx, WI, USA) were deployed at 130 locations across the landscape between May and September 2015 (Figure ). These locations were separated by a mean distance of 1.4 km and distributed across an elevational gradient (M = 376 m.a.s.l.; range = 64–735 m.a.s.l.). Accounting for theft, vandalism and malfunction, data were retrieved from 121 locations. We stratified our sampling according to HCS strata, while capturing the broader heterogeneity of the landscape using reference classes (protected “Continuous Logged Forest” and well‐established “Oil Palm Plantation”) for comparative purposes. As the extent of Scrub and Open Land was relatively low compared to the other classes, these strata were pooled into a single class, “Developed Land”, for biodiversity analyses: Continuous Logged Forest, N = 27; Dense Forest, N = 23; Young Regenerating Forest, N = 16; Developed Land, N = 26; and, Oil Palm Plantation, N = 23.Due to the number of cameras available, data collection was completed over two rotations, each comprising 65 locations. Single units were deployed for 42 consecutive nights per location, yielding a total survey effort of 4,669 camera nights. Cameras were positioned at a standardised height of 30 cm, on low resistance travel routes (e.g. riparian areas, logging roads, skid trails) and off‐trail to account for inter and intraspecific differences in habitat use.Prior to analyses, all images that could not be identified to species level were discarded (blurred images and photos of non‐target species, equating to 17.6% of 142,294 images). Species encounters were considered independent events if they contained different individuals or were separated by a period of >60 min. A detection matrix was developed for each species, whereby 42‐day sampling periods were divided into six, 7‐day temporal replicates. Any camera site active for fewer than seven days was excluded from analysis, leaving 115 analytical units each with 2–6 replicates.Modelling frameworkWe employed hierarchical Bayesian multispecies occupancy modelling (Dorazio & Royle, ) to estimate species diversity from camera data. Hierarchical models permit the separation of ecological and sampling processes that may influence the data (Gelman & Hill, ). In the context of occupancy, this means that true absences can be differentiated from non‐detection by explicitly defining models for occurrence and detection.Multispecies occupancy models take single‐species occupancy detection models as building units (Guillera‐Arroita, ). Following Zipkin, Royle, Dawson, and Bates (), we denote the occurrence of species i at site j by the binary variable zi,j (1 = species presence; 0 = species not detected). The occurrence state is described as the outcome of a Bernoulli process, zi,j ~ Bern(ψi,j), where ψi,j denotes the occurrence probability. The true occurrence state is imperfectly observed, so the model includes a second Bernoulli process, xi,j,k ~ Bern(pi,j,k · zi,j), where xi,j,k is the observed detection/non‐detection data, k is the survey replicate and pi,j,k represents the corresponding detection probability conditional to species presence. The product pi,j,k · zi,j reflects that detection at sites where the species is present (zi,j = 1) happens with detection probability pi,j,k, and that detection is not possible at sites where the species is absent (zi,j = 0). We assume that variation in the abundance of a species across sampling sites does not affect species detection probabilities pi,j,k (Royle & Dorazio, ).Occurrence and detection models for individual species were linked via a hierarchical component that modelled regression coefficients as realisations from a common community‐level distribution with (hyper)parameters. Under this approach, species are assumed to respond to environmental conditions in a similar, but not identical, manner. Derived species estimates are, therefore, a compromise between individual response and the average response of the community. This results in shrinkage (the borrowing of information by individuals across the community), which has been shown to improve estimation precision, particularly for rare or elusive species that are infrequently detected during surveys (Pacifici, Zipkin, Collazo, Irizarry, & DeWan, ). We report (hyper)parameters to provide an indication of community‐level responses to covariates.Spatial concordance between HCS classes, carbon and biodiversityTo assess how mammal community representation could vary according to policy relevant carbon variables, we described occupancy and detectability using three models:Model 1logit(ψi,j) ═ μ(i)HCS Class(j)logit(pi,j,k) ═ υ(i)HCS Class(j)Model 2logit(ψi,j) ═ μi + α1iCC1000j + α2iCC10002jlogit(pi,j,k) ═ υ(i)HCS Class(j)Model 3logit(ψi,j) ═ μi+ α1iCC25j + α2iCC252jlogit(pi,j,k) ═ υ(i)HCS Class(j)Occupancy and detection probabilities were modelled with intercepts on the logit scale, specific for each species and HCS class (Model 1). Continuous measures of carbon, including quadratic terms, were incorporated into occurrence models alongside species‐specific intercepts to determine the potential for REDD+ to deliver biodiversity co‐benefits (Models 2 and 3). These carbon data were from two sources: coarse‐grained 1 km resolution global maps (“CC1000”; Avitabile et al., ), and 25 m resolution maps derived from biomass estimates from the study site linked to Rapideye™ satellite imagery (“CC25”; Pfeifer et al., ; for a subset of sites not obscured by cloud cover, N = 66). HCS‐specific intercepts were retained in the detection components of Models 2 and 3 as they broadly describe the influence of habitat type. We chose to model HCS, CC1000 and CC25 separately due to strong evidence of collinearity between these variables (|r| ≥ 0.7). Continuous carbon and HCS covariates were calculated as average values extracted from a 100 m buffer (c. 3.1 ha area) around each camera location. Covariates were centred and standardised prior to analysis. We found no evidence of spatial autocorrelation in the detection dataset (Moran's I = 0.08 ≤ p ≤ .92), indicating that assumptions of independence in occupancy modelling were met (Royle & Dorazio, ).The models were fitted to include inference about the number of potential species not observed during sampling (Dorazio & Royle, ). To achieve this, detection data were augmented with 50 hypothetical species, with all‐zero encounter histories, following Royle, Dorazio, and Link (). Predicted species richness was calculated for each camera location allowing for post hoc comparison between HCS classes.We compared mammal richness between HCS classes using a Bayesian linear model. We follow a two‐stage analytical approach described by Kéry and Royle (), whereby estimation uncertainty associated with predicted species richness is propagated by the inclusion of an additional residual component into the model (SD of richness estimates from the hierarchical Bayesian multispecies occupancy models). In principle, parameter estimates could be derived directly from a single model, but this resulted in lower precision. Since land‐use change disproportionately affects species of conservation concern and disturbance‐sensitive forest specialists, we report our findings for: (1) all species; (2) threatened species (IUCN red‐listed as vulnerable, endangered or critically endangered); (3) non‐threatened species (IUCN least concern or near‐threatened); (4) disturbance‐sensitive species (listed as medium‐high sensitivity according to Wilson et al., ), and; (5) disturbance‐tolerant species (low sensitivity; see Appendix S3 for species‐specific group assignment).Biodiversity co‐benefits of REDD+To assess the potential biodiversity co‐benefits of REDD+, we extracted predicted species richness values from the hierarchical occurrence model and explored their association with carbon. To determine if these relationships were grain‐dependent, we derived carbon data from coarse‐ (CC1000, 1 km) and fine‐grained (CC25, 25 m) satellite‐derived datasets. Associations between levels of mammal species richness and carbon at the two different spatial resolutions, and for a priori groupings, were assessed via Bayesian two‐stage linear models incorporating quadratic terms.All analyses were conducted in WinBUGS version 1.4.3 through r version 3.3.0 using the package “R2WinBUGS” (Sturtz, Ligges, & Gelman, ); see Appendix S4 for further information on model specification and predictive performance checks. A list of datasets used for analysis is provided in the Data Sources section.RESULTSCamera‐trapping yielded 3,237 independent capture events of 28 species, comprising 24 genera distributed across 16 families. In contrast, our models predicted 30.6 species across the landscape (95% Bayesian Credible Interval, BCI = 28.0–37.0), suggesting that few mammal species were missed by our sampling. The effect of imperfect detection was more pronounced at the camera‐trap level, where predicted richness was consistently greater than observed richness (M = 4.35, range = 0.02–12.26).Spatial concordance between biodiversity and HCS classesHierarchical Bayesian multispecies models indicated reduced mammalian occupancy in the low carbon strata (Figure ). Community (hyper)parameters revealed comparable estimates of mean occupancy between Continuous Logged Forest (M = 0.49, BCI = 0.32–0.63), Dense Forest (0.36, 0.17–0.60) and Developed Land (0.32, 0.12–0.56). However, community occupancy was low in Young Regenerating Forest (0.23, 0.11–0.45) and Oil Palm plantation (0.05, 0.01–0.31).Caterpillar plots of outputs from the hierarchical Bayesian multispecies occupancy model. Graphs show species‐specific baseline occupancy estimates (including 95% Bayesian credible interval) relative to habitat class (a–e). Mean community (hyper)parameter occupancy values and their associated credible intervals are represented in the shaded (orange) background to each plot. Species exhibiting deviations from a baseline occupancy of 0.5 are shown with shaded (blue) bars [Colour figure can be viewed at wileyonlinelibrary.com]Our models demonstrated species‐specific associations with HCS classes (Figure ). For example, occupancy estimates indicate that Sus barbatus Müller (bearded pig) and Macaca nemestrina Linnaeus (southern pig‐tailed macaque) were common in Continuous Logged Forest (S. barbatus: 0.71, 0.53–0.85; M. nemestrina: 0.71, 0.53‐0.86) and Dense Forest (S. barbatus: 0.74, 0.53–0.90; M. nemestrina: 0.74, 0.52–0.92), with occupancy of M. nemestrina also high in Developed Land (0.71, 0.51–0.87). Conversely, species such as Tragulus kanchil Raffles (lesser mouse‐deer: 0.20, 0.08–0.40) and Helarctos malayanus Raffles (sun bear: 0.21, 0.08–0.44) were rare in Dense Forest. In the Oil Palm plantation five species demonstrated low occupancy, four of which were threatened taxa (Figure e). Species‐specific detection summaries for the HCS model are available in Appendix S5.3.Extremes in predicted species richness were identified between the reference habitat classes (Figure ); Continuous Logged Forest was found to have the highest richness (14.12, 13.20–15.07), while Oil Palm plantation supported the most depauperate community (4.54, 3.58–5.52). Estimates of total richness were similar between Dense Forest (11.38, 10.30–12.51) and Developed Land (10.63, 9.52–11.02), while the number of species found in Young Regenerating Forest was significantly lower (8.15, 7.13–9.27). These patterns were consistent across groupings.Boxplots demonstrating species richness in relation to habitat class for: (1) all species; (2) threatened species (IUCN Red Listed as vulnerable, endangered or critically endangered); (3) disturbance‐sensitive species. Letters indicate significant differences between habitat classes within broader species groupings (different letters suggest significance while identical letters indicate non‐significance) [Colour figure can be viewed at wileyonlinelibrary.com]Biodiversity co‐benefits of REDD+The global‐ vs. local‐scale carbon values at camera locations were inconsistent. The 1 km resolution global data tended to produce much higher carbon estimates compared to those derived from higher resolution imagery (global mean = 152.23 t C/ha, range = 50.39–236.53; local mean = 22.95 t C/ha, range = 0.31–94.98). Carbon values from the global‐ and local‐scale maps corresponded broadly with biomass values derived from field inventories (N = 164; rs = 0.55 global; rs = 0.51; local‐scale). However, local‐scale carbon estimates were found to be more precise (RMSE: local = 29.05 t C/ha; global = 130.94 t C/ha). We found no influence of continuous measures of carbon on mammalian occupancy using either global‐ or local‐scale carbon data (see Appendix S5.4/S5.6). Species‐specific detection summaries for the continuous carbon models are available in Appendix S5.5/S5.7.Grain‐dependency between the association of carbon and mammal richness was evident. Using global carbon data, no relationship between the two variables was apparent, regardless of the species grouping (Figure a,c,e). However, at the local‐scale, positive associations with carbon were identified for threatened and disturbance‐sensitive species (Figure d,f). This trend was not consistent across groupings with all species, non‐threatened and disturbance‐tolerant taxa demonstrating no relationship with carbon (Figure b).Bayesian linear model outputs demonstrating significant positive relationships between predicted species richness and carbon stock estimates derived from a 25 m resolution local dataset (d: threatened species; f: disturbance‐sensitive species). All other associations presented were found to be non‐significant. Solid (blue) lines indicate predicted mean posterior distribution values, dashed lines refer to predicted 95% Bayesian credible intervals and vertical grey lines highlight the error associated with each estimated species richness value [Colour figure can be viewed at wileyonlinelibrary.com]DISCUSSIONThe extent to which biodiversity and carbon spatially align is fundamental to our understanding of whether carbon‐based policies can deliver positive results for conservation in human‐modified landscapes. Among the few studies that assess biodiversity and carbon covariance using primary and/or high‐resolution data (Magnago et al., ; Sollmann et al., ), ours is the first to verify an association within a tropical landscape mosaic undergoing certification. We show that the strength, nature and extent of biodiversity co‐benefits are dependent on how carbon stocks are characterised (i.e. categorical or continuous), the spatial resolution of the carbon data employed, and the species considered.Contribution of the HCS approach to biodiversity conservationWhen evaluating community‐level responses to HCS classes, we found comparable levels of mammalian occupancy between Continuous Logged Forest, Developed Land and Dense Forest, while occupancy was reduced in Young Regenerating Forest and Oil Palm. Occupancy can be a viable surrogate for abundance under certain conditions (Efford & Dawson, ). Our results could therefore suggest the persistence of certain mammal species at lower densities within carbon‐poor classes, which confirms previous reports of reduced mammalian abundance in impoverished forest habitats (Bicknell, Struebig, Edwards, & Davies, ).Occupancy and species richness estimates for the total mammal community highlight comparable levels of biodiversity between the Dense Forest and Developed Land classes, supporting previous studies that demonstrate the conservation value of heavily degraded forest for a range of taxonomic groups (Edwards et al., ; Struebig et al., ; Wearn, Carbone, Rowcliffe, Bernard, & Ewers, ). However, we advise caution when interpreting the biodiversity value of Developed Land, which may not be fully realised for long‐lived mammal species until extinction debts, owed to a legacy of disturbance, are repaid (Rosa, Smith, Wearn, Purves, & Ewers, ). The biodiversity value of Developed Land that we found is also crucially dependent on the low levels of hunting at our study site. Hunting has been shown to have substantial impacts on mammal communities elsewhere in the region (Harrison et al., ). Our study adds to the growing body of evidence that shows oil palm plantations to have depauperate mammalian communities, comprised of few generalist species occurring at low densities (Wearn et al., ; Yue, Brodie, Zipkin, & Bernard, ). While our data generally conform to the pattern of declining biodiversity relative to structural complexity, Young Regenerating Forest demonstrated comparably lower biodiversity value regardless of the metric examined. Given the difficulties in differentiating between the Young Regenerating Forest and Scrub strata (see Appendix S2), we believe this finding reflects ambiguities in the HCS classification process.While our analyses demonstrate differences in conservation value between the HCS strata, the ability of land parcels to support biodiversity will also be limited by habitat fragmentation effects. This process is pervasive in human‐modified landscapes, and has contributed to species richness declines of up to 75% (Haddad et al., ). While efforts to account for habitat fragmentation in the HCS prioritisation process are underway, a definitive toolkit is still in development. Therefore, while we have not explicitly accounted for the independent and interactive effects of fragmentation metrics (e.g. patch size, isolation and connectivity) on biodiversity in our analyses, it warrants further consideration as the HCS Approach gains traction across the agricultural sector.Contribution of REDD+ to biodiversity conservationOur results indicate that spatial concordance between biodiversity and carbon can be overlooked if the latter is calculated via low‐resolution data. Using carbon information from a commonly utilised global dataset, no association with mammal diversity was identified, suggesting that REDD+ initiatives would not provide biodiversity co‐benefits in heavily degraded landscapes. However, when high‐resolution carbon maps were employed, a positive relationship with species richness was found for threatened and disturbance‐sensitive taxa, demonstrating the value of REDD+ to those species most vulnerable to land‐use change. When all species were considered, these relationships were obscured by non‐threatened, generalist species that are resilient to disturbance. Our findings provide further support for biodiversity co‐benefits in agricultural land‐use mosaics, as previously demonstrated for a range of taxonomic groups (birds and dung beetles: Gilroy et al., ; amphibians: Basham et al., ), while highlighting important nuances in the carbon–biodiversity relationship. We advocate the use of fine‐grained, field‐validated carbon data when determining the extent and nature of biodiversity co‐benefits and suggest an emphasis on species of conservation concern.Our detailed landscape appraisal is the first to identify biodiversity co‐benefits for mammals, a taxonomic group that occupies key trophic positions in tropical forest ecosystems and is frequently prioritised by conservation. Previous studies have proved less convincing. Across a pantropical network of sites, Beaudrot et al. () found no association between forest carbon and three measures of mammalian diversity. However, by aggregating fine‐scale biomass data at the site level, the authors compromised the resolution of their data, potentially obscuring intra‐site relationships that would be more representative of a REDD+ management unit. Similarly, Sollmann et al. () found little correspondence between above‐ground biomass and mammal occupancy in a certified forest reserve in Malaysian Borneo, despite adopting a comparable methodology to the present study. Contrasting findings may be attributed to spatial variability in hunting pressure.Implications for HCS implementationGiven that RSPO members have little obligation to protect highly disturbed forest of uncertain conservation value, the HCS Approach is a useful tool to designate high carbon, high biodiversity land in areas that would otherwise be converted to plantation. HCS areas can also contribute to national and regional spatial planning initiatives that mitigate the effects of environmental change on tropical biodiversity by promoting connectivity in human‐modified landscapes (Struebig et al., ).Under current HCS guidelines, 62% (8,150 ha) of the remaining forest in the study system would qualify for protection from agricultural conversion, equating to a net gain of 15.72 t C/ha (see Appendix S2) at an annual opportunity cost of US$3.7 million (based on Fisher et al., ). The success of certification depends on financial returns from sustainable production offsetting the economic losses associated with sustainable practices. While the zero deforestation principle of the HCS Approach reduces reputational risk by aligning with consumer goods forum calls to eliminate deforestation from global commodity supply chains, it has been considered economically restrictive for nations with extensive pristine forests (Senior, Brown, Villalpando, & Hill, ), indicating that current guidelines may be too stringent. Strata such as Young Regenerating Forest might, therefore, end up being earmarked for conversion rather than conservation in some circumstances. However, with the conservation value of this stratum likely to increase as forests regenerate, the impact of such a policy change needs to be fully evaluated. Carbon neutral conversion represents an alternative to the current emphasis on zero deforestation. While the specific carbon threshold for delineating forest has proved contentious, Pirker et al. () demonstrated that protecting areas exceeding 100 t C/ha would safeguard 73% of the climatically suitable area for oil palm expansion. Ultimately, compromise begets progression, and while the industry should still strive for zero deforestation, carbon neutral conversion may be more viable in specific countries and circumstances, if agricultural expansion, economic development and forest conservation are to be reconciled.Implications for REDD+ implementationThe considerable enthusiasm for biodiversity co‐benefits often obscures the fact that REDD+ is fundamentally a carbon‐orientated mechanism with limited scope for increasing biodiversity conservation (Venter, Hovani, Bode, & Possingham, ). While we provide further evidence to verify biodiversity co‐benefits in human‐modified landscapes, it is unlikely that REDD+ will be economically viable in carbon‐poor environments. Given current economic pressures and weak carbon markets, REDD+ projects currently prioritise carbon gains at low operating costs. Acting optimally for carbon will therefore place increasing agricultural pressure on secondary or degraded forests that are comparatively low in carbon value but retain appreciable levels of biodiversity (Edwards et al.,). Conservationists must ensure that safeguards are in place to support vulnerable species in disturbed habitats that fall beyond the remit of carbon‐financing mechanisms.The viability of REDD+ in human‐modified landscapes is further hindered by the profitability of oil palm. Under current voluntary markets, avoided deforestation through REDD+ was found to have an opportunity cost of $3,221–8,636 ha−1 over a 30‐year period when compared to potential profits generated from oil palm (Butler et al., ). For REDD+ to be an economically competitive alternative to oil palm cultivation, climate change policies must legitimise REDD+ carbon credits to facilitate their trade on financially lucrative compliance markets (Butler et al., ).CONCLUSIONSOur work highlights the potential for environmental certification and REDD+ financing mechanisms to work in tandem with conservation to mitigate the effects of agricultural expansion on tropical forest carbon stocks and biodiversity. REDD+ is well placed if it prioritises large tracts of contiguous forest, especially if commitments to carbon stock enhancement safeguard degraded forest of biological value. Certification schemes, coupled with land‐use planning tools such as HCS, can help secure sizeable forest patches of high conservation value in agricultural estates, and offer a further safeguard to minimise encroachment. Conservationists should capitalise on both types of carbon‐based policy to maximise the potential for developed lands to provide ecological stepping stones for threatened wildlife between a network of high carbon, high biodiversity areas.ACKNOWLEDGEMENTSThis study was funded by the UK Natural Environment Research Council (NERC) (NE/K016407/1; http://lombok.hmtf.info/) and a NERC EnvEast PhD studentship to N.J.D. G.G.A. is the recipient of a Discovery Early Career Research Award from the Australian Research Council (DE160100904). We thank the Sabah Biodiversity Council, Sabah Forest Department, Yayasan Sabah, Sime Darby and Benta Wawasan for permitting access. We are indebted to Jamiluddin Jami, Arnold James, Mohd. Mustamin, Ampat Siliwong, Sabidee Mohd. Rizan, Najmuddin Jamal and Lois Kinneen for field assistance. We also thank Rob McWilliam, Tri Agus Sugiyanto (both The Forest Trust), Grant Rosoman and Sapta Ananda (both Greenpeace) for training on HCS techniques.AUTHORS' CONTRIBUTIONSN.J.D., M.J.S. and Z.G.D. conceived the ideas and designed the methodology; N.J.D. and E.L.B. collected and processed data, adapting scripts developed by O.R.W.; M.P. and N.J.D. provided carbon data; N.J.D. and G.G.A. undertook analyses; N.J.D., M.J.S. and Z.G.D. led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.DATA ACCESSIBILITYData available from the NERC Environmental Information Data Centre following an embargo period (accessible from: 31/08/2019). Detection data for 28 observed mammal species: https://doi.org/10.5285/62774180-ae72-4873-9482-e8be3935f533 (Deere, Guillera‐Arroita, & Baking, ); HCS stratification of the study site: https://doi.org/10.5285/81cad1ef-b5cc-4592-a71f-204a5d04b700 (Deere et al., ).REFERENCESAvitabile, V., Herold, M., Heuvelink, G. B. M., Lewis, S. L., Phillips, O. L., Asner, G. P., … Willcock, S. (2016). An integrated pan‐tropical biomass map using multiple reference datasets. Global Change Biology, 22, 1406–1420. https://doi.org/10.1111/gcb.2016.22.issue-4Baccini, A., Goetz, S. J., Walker, W. S., Laporte, N. T., Sun, M., Sulla‐Menashe, D., … Houghton, R. A. (2012). Estimated carbon dioxide emissions from tropical deforestation improved by carbon‐density maps. Nature Climate Change, 2, 182–185. https://doi.org/10.1038/nclimate1354Basham, E. W., González del Pliego, P., Acosta‐Galvis, A. R., Woodcock, P., Medina Uribe, C. A., Haugaasen, T, … Edwards, D. P. (2016). Quantifying carbon and amphibian co‐benefits from secondary forest regeneration in the Tropical Andes. Animal Conservation, 19, 548–560. https://doi.org/10.1111/acv.12276Beaudrot, L., Kroetz, K., Alvarez‐Loayza, P., Amaral, L., Breuer, T., Fletcher, C., … Andelman, S. (2016). Limited carbon and biodiversity co‐benefits for tropical forest mammals and birds. Ecological Applications, 26, 1098–1111. https://doi.org/10.1890/15-0935Bicknell, J. E., Struebig, M. J., Edwards, D. P., & Davies, Z. G. (2014). Improved timber harvest techniques maintain biodiversity in tropical forests. Current Biology, 24, R1119–R1120. https://doi.org/10.1016/j.cub.2014.10.067Butler, R. A., Koh, L. P., & Ghazoul, J. (2009). REDD in the red: Palm oil could undermine carbon payment schemes. Conservation Letters, 2, 67–73. https://doi.org/10.1111/conl.2009.2.issue-2Chave, J., Rejou‐Mechain, M., Burquez, A., Chidumayo, E., Colgan, M. S., Delitti, W. B. C., … Vieilledent, G (2014). Improved allometric models to estimate the aboveground biomass of tropical trees. Global Change Biology, 20, 3177–3190. https://doi.org/10.1111/gcb.2014.20.issue-10Deere, N. J., Guillera‐Arroita, G., Baking, E. L., Bernard, H., Pfeifer, M., Reynolds, G., … Struebig, M. J. (2017a). Mammal detection data for the SAFE project site, Sabah, Malaysian Borneo, 2015 [HTMF]. NERC Environmental Information Data Centre. https://doi.org/10.5285/62774180-ae72-4873-9482-e8be3935f533Deere, N. J., Guillera‐Arroita, G., Baking, E. L., Bernard, H., Pfeifer, M., Reynolds, G., … Struebig, M. J. (2017b). High Carbon Stock stratification of the SAFE project site, Sabah, Malaysian Borneo, 2015 [HTMF]. NERC Environmental Information Data Centre. https://doi.org/10.5285/81cad1ef-b5cc-4592-a71f-204a5d04b700Dorazio, R. M., & Royle, J. A. (2005). Estimating size and composition of biological communities by modeling the occurrence of species. Journal of the American Statistical Association, 100, 389–398. https://doi.org/10.1198/016214505000000015Edwards, D. P., Fisher, B., & Wilcove, D. S. (2012). High conservation value or high confusion value? Sustainable agriculture and biodiversity conservation in the tropics. Conservation Letters, 5, 20–27. https://doi.org/10.1111/conl.2012.5.issue-1Edwards, D. P., Magrach, A., Woodcock, P., Ji, Y., Lim, N. T.‐L., Edwards, F. A., … Yu, D. W. (2014). Selective‐logging and oil palm: Multitaxon impacts, biodiversity indicators, and trade‐offs for conservation planning. Ecological Applications, 24, 2029–2049. https://doi.org/10.1890/14-0010.1Efford, M. G., & Dawson, D. K. (2012). Occupancy in continuous habitat. Ecosphere, 3, 1–15.Egoh, B., Reyers, B., Rouget, M., Bode, M., & Richardson, D. M. (2009). Spatial congruence between biodiversity and ecosystem services in South Africa. Biological Conservation, 142, 553–562. https://doi.org/10.1016/j.biocon.2008.11.009Ewers, R. M., Didham, R. K., Fahrig, L., Ferraz, G., Hector, A., Holt, R. D., … Turner, E. C. (2011). A large‐scale forest fragmentation experiment: The stability of altered forest ecosystems project. Philosophical Transactions of the Royal Society B‐Biological Sciences, 366, 3292–3302. https://doi.org/10.1098/rstb.2011.0049Fisher, B., Edwards, D. P., Giam, X. L., & Wilcove, D. S. (2011). The high costs of conserving Southeast Asia's lowland rainforests. Frontiers in Ecology and the Environment, 9, 329–334. https://doi.org/10.1890/100079Fitzherbert, E. B., Struebig, M. J., Morel, A., Danielsen, F., Bruhl, C. A., Donald, P. F., & Phalan, B. (2008). How will oil palm expansion affect biodiversity? Trends in Ecology & Evolution, 23, 538–545. https://doi.org/10.1016/j.tree.2008.06.012Gardner, T. A., Barlow, J., Araujo, I. S., Ávila‐Pires, T. C., Bonaldo, A. B., Costa, J. E., … Peres, C. A. (2008). The cost‐effectiveness of biodiversity surveys in tropical forests. Ecology Letters, 11, 139–150. https://doi.org/10.1111/j.1461-0248.2007.01133.xGardner, T. A., Burgess, N. D., Aguilar‐Amuchastegui, N., Barlow, J., Berenguer, E., Clements, T., … Vieira, I. C. G. T. (2012). A framework for integrating biodiversity concerns into national REDD+ programmes. Biological Conservation, 154, 61–71. https://doi.org/10.1016/j.biocon.2011.11.018Gaveau, D. L. A., Sloan, S., Molidena, E., Yaen, H., Sheil, D., Abram, N. K., … Meijaard, E. (2014). Four decades of forest persistence, clearance and logging on borneo. PLoS ONE, 9, e101654. https://doi.org/10.1371/journal.pone.0101654Gelman, A., & Hill, J. (2007). Data analysis using regression and multilevel/hierarchical models. Cambridge, UK: Cambridge University Press.Gibbs, H. K., Ruesch, A. S., Achard, F., Clayton, M. K., Holmgren, P., Ramankutty, N., & Foley, J. A. (2010). Tropical forests were the primary sources of new agricultural land in the 1980s and 1990s. Proceedings of the National Academy of Sciences of the United States of America, 107, 16732–16737. https://doi.org/10.1073/pnas.0910275107Gillies, C. S., & St Clair, C. C. (2010). Functional responses in habitat selection by tropical birds moving through fragmented forest. Journal of Applied Ecology, 47, 182–190. https://doi.org/10.1111/jpe.2010.47.issue-1Gilroy, J. J., Woodcock, P., Edwards, F. A., Wheeler, C., Baptiste, B. L. G., Medina Uribe, C. A, … Edwards, D. P. (2014). Cheap carbon and biodiversity co‐benefits from forest regeneration in a hotspot of endemism. Nature Climate Change, 4, 503–507. https://doi.org/10.1038/nclimate2200Guillera‐Arroita, G. (2017). Modelling of species distributions, range dynamics and communities under imperfect detection: Advances, challenges and opportunities. Ecography, 40, 281–295. https://doi.org/10.1111/ecog.02445Haddad, N. M., Brudvig, L. A., Clobert, J., Davies, K. F., Gonzalez, A., & Holt, R. D, … Townshend, J. R. (2015). Habitat fragmentation and its lasting impact on Earth's ecosystems. Science Advances, 1, e1500052.Hansen, M. C., Potapov, P. V., Moore, R., Hancher, M., Turubanova, S. A., Tyukavina, A, … Townshend, R. G. (2013). High‐resolution global maps of 21st‐century forest cover change. Science, 342, 850–853. https://doi.org/10.1126/science.1244693Harrison, R. D., Sreekar, R., Brodie, J. F., Brook, S., Luskin, M., O'Kelly, H, … Velho, N. (2016). Impacts of hunting on tropical forests in Southeast Asia. Conservation Biology, 30, 972–981. https://doi.org/10.1111/cobi.12785Kéry, M., & Royle, J. A. (2015). Applied hierarchical modeling in ecology: Analysis of distribution, abundance and species richness in R and BUGS: Volume 1: prelude and static models. London: Academic Press.Kessler, M., Hertel, D., Jungkunst, H. F., Kluge, J., Abrahamczyk, S., Bos, M, … Tscharntke, T. (2012). Can joint carbon and biodiversity management in tropical agroforestry landscapes be optimized? PLoS ONE, 7, e47192. https://doi.org/10.1371/journal.pone.0047192Magnago, L. F. S., Magrach, A., Laurance, W. F., Martins, S. V., Meira‐Neto, J. A. A., Simonelli, M., & Edwards, D. P. (2015). Would protecting tropical forest fragments provide carbon and biodiversity co‐benefits under REDD+? Global Change Biology, 21, 3455–3468. https://doi.org/10.1111/gcb.2015.21.issue-9Mitchard, E. T. A., Feldpausch, T. R., Brienen, R. J. W., Lopez‐Gonzalez, G., Monteagudo, A., Baker, T. R, … Phillips, O. L. (2014). Markedly divergent estimates of Amazon forest carbon density from ground plots and satellites. Global Ecology and Biogeography, 23, 935–946. https://doi.org/10.1111/geb.2014.23.issue-8Murray, J. P., Grenyer, R., Wunder, S., Raes, N., & Jones, J. P. (2015). Spatial patterns of carbon, biodiversity, deforestation threat, and REDD+ projects in Indonesia. Conservation Biology, 29, 1434–1445. https://doi.org/10.1111/cobi.12500Naidoo, R., Balmford, A., Costanza, R., Fisher, B., Green, R. E., Lehner, B. T. R, … Ricketts, T. H. (2008). Global mapping of ecosystem services and conservation priorities. Proceedings of the National Academy of Sciences of the United States of America, 105, 9495–9500. https://doi.org/10.1073/pnas.0707823105Pacifici, K., Zipkin, E. F., Collazo, J. A., Irizarry, J. I., & DeWan, A. (2014). Guidelines for a priori grouping of species in hierarchical community models. Ecology and Evolution, 4, 877–888. https://doi.org/10.1002/ece3.2014.4.issue-7Pfeifer, M., Kor, L., Nilus, R., Turner, E., Cusack, J., Lysenko, I., … Ewers, R. M. (2016b). Mapping the structure of Borneo's tropical forests across a degradation gradient. Remote Sensing of Environment, 176, 84–97. https://doi.org/10.1016/j.rse.2016.01.014Pirker, J., Mosnier, A., Kraxner, F., Havlík, P., & Obersteiner, M. (2016). What are the limits to oil palm expansion? Global Environmental Change, 40, 73–81. https://doi.org/10.1016/j.gloenvcha.2016.06.007Potter, L. (2015). Managing oil palm landscapes: A seven‐country survey of the modern palm oil industry in Southeast Asia, Latin America and West Africa. CIFOR.Rodríguez‐Castañeda, G., Hof, A. R., Jansson, R., & Harding, L. E. (2012). Predicting the fate of biodiversity using species' distribution models: Enhancing model comparability and repeatability. PLoS ONE, 7, e44402. https://doi.org/10.1371/journal.pone.0044402Rosa, I. M., Smith, M. J., Wearn, O. R., Purves, D., & Ewers, R. M. (2016). The environmental legacy of modern tropical deforestation. Current Biology, 26, 2161–2166. https://doi.org/10.1016/j.cub.2016.06.013Rosoman, G., Sheun, S. S., Opal, C., Anderson, P., & Trapshah, R. (Eds.) (2017). The HCS approach toolkit. Singapore: HCS Steering Group.Royle, J., & Dorazio, R. (2008). Hierarchical modelling and inference in ecology. New York, NY: Academic Press.Royle, J. A., Dorazio, R. M., & Link, W. A. (2007). Analysis of multinomial models with unknown index using data augmentation. Journal of Computational and Graphical Statistics, 16, 67–85. https://doi.org/10.1198/106186007X181425RSPO (2017). Certification in numbers. Retrieved from www.rspo.org/about/impacts.Ruiz‐Jaen, M. C., & Potvin, C. (2010). Tree diversity explains variation in ecosystem function in a neotropical forest in Panama. Biotropica, 42, 638–646. https://doi.org/10.1111/j.1744-7429.2010.00631.xSenior, M. J., Brown, E., Villalpando, P., & Hill, J. K. (2015). Increasing the scientific evidence base in the “high conservation value”(HCV) approach for biodiversity conservation in managed tropical landscapes. Conservation Letters, 8, 361–367. https://doi.org/10.1111/conl.12148Sollmann, R., Mohamed, A., Niedballa, J., Bender, J., Ambu, L., Lagan, P, … Wilting, A. (2017). Quantifying mammal biodiversity co‐benefits in certified tropical forests. Diversity and Distributions, 23, 317–328. https://doi.org/10.1111/ddi.2017.23.issue-3Strassburg, B. B. N., Kelly, A., Balmford, A., Davies, R. G., Gibbs, H. K., Lovett, A, … Rodrigues, A. S. L. (2010). Global congruence of carbon storage and biodiversity in terrestrial ecosystems. Conservation Letters, 3, 98–105. https://doi.org/10.1111/conl.2010.3.issue-2Struebig, M. J., Kingston, T., Petit, E. J., Le Comber, S. C., Zubaid, A., Mohd‐Adnan, A., & Rossiter, S. J. (2011). Parallel declines in species and genetic diversity in tropical forest fragments. Ecology Letters, 14, 582–590. https://doi.org/10.1111/j.1461-0248.2011.01623.xStruebig, M. J., Turner, A., Giles, E., Lasmana, F., Tollington, S., Bernard, H., & Bell, D. (2013). Quantifying the biodiversity value of repeatedly logged rainforests: Gradient and comparative approaches from Borneo. Advances in Ecological Research, 48, 183–224. https://doi.org/10.1016/B978-0-12-417199-2.00003-3Struebig, M. J., Wilting, A., Gaveau, D. L., Meijaard, E., Smith, R. J., Fischer, M., … Wong, A., (2015). Targeted conservation to safeguard a biodiversity hotspot from climate and land‐cover change. Current Biology, 25, 372–378. https://doi.org/10.1016/j.cub.2014.11.067Sturtz, S., Ligges, U., & Gelman, A. (2005). R2WinBUGS: a package for running WinBUGS from R. Journal of Statistical software, 12, 1–16.Venter, O., Hovani, L., Bode, M., & Possingham, H. (2013). Acting optimally for biodiversity in a world obsessed with REDD+. Conservation Letters, 6, 410–417. https://doi.org/10.1111/conl.2013.6.issue-6Venter, O., & Koh, L. P. (2012). Reducing emissions from deforestation and forest degradation (REDD+): Game changer or just another quick fix? Annals of the New York Academy of Sciences, 1249, 137–150. https://doi.org/10.1111/j.1749-6632.2011.06306.xWearn, O. R., Carbone, C., Rowcliffe, J. M., Bernard, H., & Ewers, R. M. (2016). Grain‐dependent responses of mammalian diversity to land‐use and the implications for conservation set‐aside. Ecological Applications, 26, 1409–1420. https://doi.org/10.1890/15-1363Wilcove, D. S., Giam, X., Edwards, D. P., Fisher, B., & Koh, L. P. (2013). Navjot's nightmare revisited: Logging, agriculture, and biodiversity in Southeast Asia. Trends in Ecology & Evolution, 28, 531–540. https://doi.org/10.1016/j.tree.2013.04.005Wilson, K. A., Meijaard, E., Drummond, S., Grantham, H. S., Boitani, L., Catullo, G, … Watts, M. (2010). Conserving biodiversity in production landscapes. Ecological Applications, 20, 1721–1732. https://doi.org/10.1890/09-1051.1Wulder, M. A., Franklin, S. E., White, J. C., Cranny, M. M., & Dechka, J. A. (2004). Inclusion of topographic variables in an unsupervised classification of satellite imagery. Canadian Journal of Remote Sensing, 30, 137–149. https://doi.org/10.5589/m03-063Yaap, B., Struebig, M. J., Paoli, G., & Koh, L. P. (2010). Mitigating the biodiversity impacts of oil palm development. CAB Reviews, 5, 1–11.Yue, S., Brodie, J. F., Zipkin, E. F., & Bernard, H. (2015). Oil palm plantations fail to support mammal diversity. Ecological Applications, 25, 2285–2292. https://doi.org/10.1890/14-1928.1Zipkin, E. F., Royle, J. A., Dawson, D. K., & Bates, S. (2010). Multi‐species occurrence models to evaluate the effects of conservation and management actions. Biological Conservation, 143, 479–484. https://doi.org/10.1016/j.biocon.2009.11.016DATA SOURCESAvitabile, V., Herold, M., Heuvelink, G. B. M., Lewis, S. L., Phillips, O. L., Asner, G. P., … Willcock, S. (2016). An integrated pan‐tropical biomass map using multiple reference datasets. Global Change Biology, 22, 1406–1420. https://doi.org/10.1111/gcb.2016.22.issue-4Pfeifer, M., Kor, L., Nilus, R., Turner, E., Cusack, J., Lysenko, I., … Ewers, R. M. (2016a). Mapping the structure of Borneo's tropical forests across a degradation gradient. Remote Sensing of Environment, 176, 84–97. https://doi.org/10.1016/j.rse.2016.01.014 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Applied Ecology Wiley

Loading next page...
 
/lp/wiley/high-carbon-stock-forests-provide-co-benefits-for-tropical-G0wMJPRAwW

References (61)

Publisher
Wiley
Copyright
Journal of Applied Ecology © 2018 British Ecological Society
ISSN
0021-8901
eISSN
1365-2664
DOI
10.1111/1365-2664.13023
Publisher site
See Article on Publisher Site

Abstract

INTRODUCTIONAgricultural expansion has emerged as a pervasive threat to tropical forests and biodiversity (Wilcove, Giam, Edwards, Fisher, & Koh, ), and has been implicated in the loss of c. 150 million ha of tropical forest over the last three decades (Gibbs et al., ; Hansen et al., ). A key driver of recent deforestation has been rising demand for cheap vegetable oil such as that from oil palm (Elaeis guineensis), which now covers 16 million ha across 43 countries, often at the expense of tropical forest (Pirker, Mosnier, Kraxner, Havlík, & Obersteiner, ).The potential economic and social benefits associated with oil palm (Potter, ) contrast with severe and well‐documented ecological impacts. Conversion of forest to oil palm plantation results in major biodiversity decline, which disproportionately affects forest specialists and species of conservation concern, resulting in assemblages dominated by disturbance‐tolerant generalists (Fitzherbert et al., ; Yaap, Struebig, Paoli, & Koh, ). With around 19% of land suitable for oil palm coinciding with areas of high biodiversity (Pirker et al., ), across forested Asia, Africa and South America, the full ecological impact of this commodity crop is yet to be fully realised. Mitigation measures that reconcile environmental sustainability, biodiversity conservation and production of crops such as oil palm are therefore essential in tropical regions.Retaining native habitat in oil palm estates is known to enhance the biological value of plantation landscapes by providing ecological refugia and improved connectivity (Gillies & St Clair, ; Struebig et al., ). However, in practice, the designation of conservation set‐asides can be hindered by agricultural profitability, with income exceeding US$11,240/ha over a 25 year growing cycle (Fisher, Edwards, Giam, & Wilcove, ). Thus, conservation efforts seeking to preserve forest within plantations may be more successful when economic incentives are provided to offset the opportunity costs associated with foregoing development. Amongst several mitigation tools available, two incentive‐driven policies based on carbon stocks have gained traction in tropical regions: (1) REDD+ (United Nations Reducing Emissions from Deforestation and forest Degradation) and related carbon credit schemes, and (2) improved land‐use planning via commodity certification (Yaap et al., ).REDD+, a payment for ecosystem services tool to mitigate climate change, aims to compensate stakeholders in developing nations for conservation initiatives and sustainable management practices that protect and restore the carbon sequestered by forests (Venter & Koh, ). If REDD+ were to achieve its economic potential, payments generated could make forest conservation financially competitive compared to oil palm cultivation (Butler, Koh, & Ghazoul, ). REDD+ is also attractive to conservation because it may deliver co‐benefits, whereby safeguarding high carbon areas also protects biodiversity at no additional cost (Gardner et al., ). However, this assumes spatial congruence between areas of high carbon and biodiversity. In reality, it is difficult to generalise on the nature, strength and extent of these co‐benefits because outcomes vary both within and between spatial scales (e.g. global: Naidoo et al., vs. Strassburg et al., ; national: Egoh, Reyers, Rouget, Bode, & Richardson, vs. Murray, Grenyer, Wunder, Raes, & Jones, ; landscape: Ruiz‐Jaen & Potvin, vs. Kessler et al., ). The extent to which carbon‐biodiversity co‐benefit assumptions hold at management unit scales appropriate to decision‐makers remains an open question.REDD+ is largely implemented at sub‐national levels. While an increasing number of studies are recognising the importance of fine‐scale assessments (e.g. Beaudrot et al., ; Magnago et al., ; Sollmann et al., ), most information on biodiversity co‐benefits is derived from global‐ and national‐scale studies that demonstrate overreliance on coarse‐grained, secondary data sources. Carbon data are typically derived from global maps (e.g. Avitabile et al., ; Baccini et al., ), which have limited application at local‐scales pertinent to management (Mitchard et al., ). Furthermore, field‐based species data are widely underrepresented in the co‐benefits literature due to the costs associated with biodiversity surveys in the tropics (Gardner et al., ). Researchers predominantly rely on coarse species range delineations, which are fraught with uncertainty (Rodríguez‐Castañeda, Hof, Jansson, & Harding, ) and may not account for localised extirpation due to anthropogenic pressure (Harrison et al., ). Despite statistical advances that account for imperfect detection in biodiversity indices (Royle & Dorazio, ), these methods have received relatively limited application in a co‐benefits context (but see Gilroy et al., ; Sollmann et al., ), resulting in possible underestimates of species assemblages. Consequently, biodiversity co‐benefits assessments at local‐scales, using primary, fine‐grained data would provide valuable policy insights.While the potential importance of REDD+ cannot be overstated, agricultural certification schemes show promise to ensure sustainable practices as companies benefit from greater access to environmentally conscious markets and increased price premiums of certified products (Yaap et al., ). The Roundtable on Sustainable Palm Oil (RSPO) is often seen as an exemplar scheme within the agricultural sector, currently certifying 21% of the global palm oil market across 2.48 million ha of land (RSPO, ). RSPO certification prohibits the conversion of high conservation value habitat in oil palm estates. However, associated assessment procedures have attracted criticism, raising concerns that current methodologies do not afford adequate biodiversity protection (Edwards, Fisher, & Wilcove, ; Yaap et al., ).The High Carbon Stock (HCS) approach has emerged as a land‐use planning tool to demarcate conservation priority areas based on carbon value, and is being explored within the RSPO architecture and that of other certification schemes. The HCS methodology seeks to conserve biodiverse and ecologically functional forest networks within agricultural concessions by directing conversion towards heavily degraded land of low carbon value (Rosoman, Sheun, Opal, Anderson, & Trapshah, ). This is achieved by stratifying land into discrete classes according to vegetation density and structure, which are then adopted as proxies for above‐ground carbon stocks and assumed to support varying levels of biodiversity. These strata are subsequently validated using field‐derived above‐ground carbon estimates, before land parcels are prioritised for conversion based on area and connectivity (Rosoman et al., ).The HCS approach has attracted widespread interest amongst agricultural industries with 10 million ha of land being evaluated across five oil palm producing countries (G. Rosoman, unpubl. data). As a model scheme, the successful integration of the HCS Approach within the RSPO framework may encourage uptake across other certifiable tropical commodities, such as rubber and soya. Nevertheless, the extent to which HCS strata correspond to areas of high biodiversity value is dependent on the accurate partitioning of vegetation classes according to their carbon value, as well as the underlying association between carbon and biodiversity. Before the HCS approach is formally adopted within certification standards, these assumptions should be tested to understand the conservation merit of the tool.Here, we determine the effectiveness of incentive‐driven carbon‐based mechanisms to safeguard biodiversity, and provide the first validation of both the carbon and biodiversity credentials of the HCS land‐use planning tool. Our appraisal focuses on a landscape undergoing conversion from forest to oil palm in Borneo, a region characterised by high deforestation and forest degradation (Gaveau et al., ; Struebig et al., ) that is typical of most HCS applications. First, we validate the accuracy of the HCS classification procedure and quantify the biodiversity value of the vegetation strata. We then assess the potential for REDD+ to deliver biodiversity co‐benefits using primary and high‐resolution data sources. To assess the influence of spatial grain on the nature of co‐benefit relationships, we compare global‐ and local‐scale measures of carbon. Throughout, we employ biodiversity indices that explicitly account for imperfect detection to provide a more accurate representation of species assemblages than simple species counts. Our work evaluates the extent to which policy options that attach greater economic significance to conservation protect vulnerable tropical forests and safeguard biodiversity.MATERIALS AND METHODSStudy systemThe study was conducted over a 13,153 ha development area comprising the Stability of Altered Forest Ecosystems project (SAFE; www.safeproject.net) and surrounding plantations in Kalabakan Forest Reserve, Sabah, Malaysian Borneo (4°46′N, 116°57′ E; Figure ). SAFE is a landscape‐scale forest modification experiment (Ewers et al., ) comprising highly disturbed lowland and hill dipterocarp forest that was logged multiple times between 1978 and 2008. The wider landscape includes near‐pristine forest in Brantian‐Tatulit Virgin Jungle Reserve, twice‐logged forest in Ulu Segama Forest Reserve, and plantations (primarily oil palm).High Carbon Stock (HCS) classification of the study landscape in Sabah, Borneo. Forest cover was delineated into four strata on the basis of vegetation density (Dense Forest, Young Regenerating Forest, Scrub, Open Land) and supplemented with two reference classes (Continuous Logged Forest, Oil Palm) to act as forest and agricultural controls. Points indicate camera‐trap locations (N = 115) [Colour figure can be viewed at wileyonlinelibrary.com]HCS classification and validationThe HCS Approach uses high‐resolution remotely‐sensed images to stratify concessions into six vegetation classes, each with unique structural characteristics (in descending order of carbon value): (1) “High Density Forest”; (2) “Medium Density Forest”; (3) “Low Density Forest”; (4) “Young Regenerating Forest”; (5) “Scrub”; and (6) “Open Land” (see Appendix S1). In practice, the High, Medium and Low Density Forest strata are aggregated as “Dense Forest” and earmarked for conservation. Young Regenerating Forest can also comprise valuable carbon stocks and is also spared from development. The threshold for allocating land for production rests on distinguishing these strata from heavily degraded Scrub and Open Land. Therefore, we mapped Dense Forest, Young Regenerating Forest, Scrub and Open Land as separate classes.All spatial data processing was implemented in ArcGIS 10.2.1 (ESRI). We used Landsat 8 and SPOT5 satellite imagery (15 and 2.5 m resolution respectively; temporal range: 2012–2014) to stratify forest habitat using HCS assessment protocols (see Appendix S1). Multiple data sources were chosen to minimise classification difficulties associated with cloud cover and haze. We undertook a supervised classification of satellite images, supplemented with visual interpretation techniques to correct for the potentially confounding effects of topographic shadow (Wulder, Franklin, White, Cranny, & Dechka, ). The resulting classes were then calibrated using above‐ground carbon values derived from forest inventory data (N = 139), collected as part of the core SAFE monitoring programme. These data conform to standardised forest inventory protocols (http://www.rainfor.org), calculating carbon as a function of above‐ground biomass (trees >10 cm DBH) using an established pantropical algorithm (Chave et al., ). Resulting HCS classes were validated using independently derived carbon estimates (Pfeifer et al., ; see Appendix S2).Camera‐trap sampling of medium‐large mammalsWe delineated terrestrial mammal diversity as these taxa are consistently prioritised in policy, land‐use planning and certification schemes. Remotely operated digital cameras (HC500 Hyperfire, Reconyx, WI, USA) were deployed at 130 locations across the landscape between May and September 2015 (Figure ). These locations were separated by a mean distance of 1.4 km and distributed across an elevational gradient (M = 376 m.a.s.l.; range = 64–735 m.a.s.l.). Accounting for theft, vandalism and malfunction, data were retrieved from 121 locations. We stratified our sampling according to HCS strata, while capturing the broader heterogeneity of the landscape using reference classes (protected “Continuous Logged Forest” and well‐established “Oil Palm Plantation”) for comparative purposes. As the extent of Scrub and Open Land was relatively low compared to the other classes, these strata were pooled into a single class, “Developed Land”, for biodiversity analyses: Continuous Logged Forest, N = 27; Dense Forest, N = 23; Young Regenerating Forest, N = 16; Developed Land, N = 26; and, Oil Palm Plantation, N = 23.Due to the number of cameras available, data collection was completed over two rotations, each comprising 65 locations. Single units were deployed for 42 consecutive nights per location, yielding a total survey effort of 4,669 camera nights. Cameras were positioned at a standardised height of 30 cm, on low resistance travel routes (e.g. riparian areas, logging roads, skid trails) and off‐trail to account for inter and intraspecific differences in habitat use.Prior to analyses, all images that could not be identified to species level were discarded (blurred images and photos of non‐target species, equating to 17.6% of 142,294 images). Species encounters were considered independent events if they contained different individuals or were separated by a period of >60 min. A detection matrix was developed for each species, whereby 42‐day sampling periods were divided into six, 7‐day temporal replicates. Any camera site active for fewer than seven days was excluded from analysis, leaving 115 analytical units each with 2–6 replicates.Modelling frameworkWe employed hierarchical Bayesian multispecies occupancy modelling (Dorazio & Royle, ) to estimate species diversity from camera data. Hierarchical models permit the separation of ecological and sampling processes that may influence the data (Gelman & Hill, ). In the context of occupancy, this means that true absences can be differentiated from non‐detection by explicitly defining models for occurrence and detection.Multispecies occupancy models take single‐species occupancy detection models as building units (Guillera‐Arroita, ). Following Zipkin, Royle, Dawson, and Bates (), we denote the occurrence of species i at site j by the binary variable zi,j (1 = species presence; 0 = species not detected). The occurrence state is described as the outcome of a Bernoulli process, zi,j ~ Bern(ψi,j), where ψi,j denotes the occurrence probability. The true occurrence state is imperfectly observed, so the model includes a second Bernoulli process, xi,j,k ~ Bern(pi,j,k · zi,j), where xi,j,k is the observed detection/non‐detection data, k is the survey replicate and pi,j,k represents the corresponding detection probability conditional to species presence. The product pi,j,k · zi,j reflects that detection at sites where the species is present (zi,j = 1) happens with detection probability pi,j,k, and that detection is not possible at sites where the species is absent (zi,j = 0). We assume that variation in the abundance of a species across sampling sites does not affect species detection probabilities pi,j,k (Royle & Dorazio, ).Occurrence and detection models for individual species were linked via a hierarchical component that modelled regression coefficients as realisations from a common community‐level distribution with (hyper)parameters. Under this approach, species are assumed to respond to environmental conditions in a similar, but not identical, manner. Derived species estimates are, therefore, a compromise between individual response and the average response of the community. This results in shrinkage (the borrowing of information by individuals across the community), which has been shown to improve estimation precision, particularly for rare or elusive species that are infrequently detected during surveys (Pacifici, Zipkin, Collazo, Irizarry, & DeWan, ). We report (hyper)parameters to provide an indication of community‐level responses to covariates.Spatial concordance between HCS classes, carbon and biodiversityTo assess how mammal community representation could vary according to policy relevant carbon variables, we described occupancy and detectability using three models:Model 1logit(ψi,j) ═ μ(i)HCS Class(j)logit(pi,j,k) ═ υ(i)HCS Class(j)Model 2logit(ψi,j) ═ μi + α1iCC1000j + α2iCC10002jlogit(pi,j,k) ═ υ(i)HCS Class(j)Model 3logit(ψi,j) ═ μi+ α1iCC25j + α2iCC252jlogit(pi,j,k) ═ υ(i)HCS Class(j)Occupancy and detection probabilities were modelled with intercepts on the logit scale, specific for each species and HCS class (Model 1). Continuous measures of carbon, including quadratic terms, were incorporated into occurrence models alongside species‐specific intercepts to determine the potential for REDD+ to deliver biodiversity co‐benefits (Models 2 and 3). These carbon data were from two sources: coarse‐grained 1 km resolution global maps (“CC1000”; Avitabile et al., ), and 25 m resolution maps derived from biomass estimates from the study site linked to Rapideye™ satellite imagery (“CC25”; Pfeifer et al., ; for a subset of sites not obscured by cloud cover, N = 66). HCS‐specific intercepts were retained in the detection components of Models 2 and 3 as they broadly describe the influence of habitat type. We chose to model HCS, CC1000 and CC25 separately due to strong evidence of collinearity between these variables (|r| ≥ 0.7). Continuous carbon and HCS covariates were calculated as average values extracted from a 100 m buffer (c. 3.1 ha area) around each camera location. Covariates were centred and standardised prior to analysis. We found no evidence of spatial autocorrelation in the detection dataset (Moran's I = 0.08 ≤ p ≤ .92), indicating that assumptions of independence in occupancy modelling were met (Royle & Dorazio, ).The models were fitted to include inference about the number of potential species not observed during sampling (Dorazio & Royle, ). To achieve this, detection data were augmented with 50 hypothetical species, with all‐zero encounter histories, following Royle, Dorazio, and Link (). Predicted species richness was calculated for each camera location allowing for post hoc comparison between HCS classes.We compared mammal richness between HCS classes using a Bayesian linear model. We follow a two‐stage analytical approach described by Kéry and Royle (), whereby estimation uncertainty associated with predicted species richness is propagated by the inclusion of an additional residual component into the model (SD of richness estimates from the hierarchical Bayesian multispecies occupancy models). In principle, parameter estimates could be derived directly from a single model, but this resulted in lower precision. Since land‐use change disproportionately affects species of conservation concern and disturbance‐sensitive forest specialists, we report our findings for: (1) all species; (2) threatened species (IUCN red‐listed as vulnerable, endangered or critically endangered); (3) non‐threatened species (IUCN least concern or near‐threatened); (4) disturbance‐sensitive species (listed as medium‐high sensitivity according to Wilson et al., ), and; (5) disturbance‐tolerant species (low sensitivity; see Appendix S3 for species‐specific group assignment).Biodiversity co‐benefits of REDD+To assess the potential biodiversity co‐benefits of REDD+, we extracted predicted species richness values from the hierarchical occurrence model and explored their association with carbon. To determine if these relationships were grain‐dependent, we derived carbon data from coarse‐ (CC1000, 1 km) and fine‐grained (CC25, 25 m) satellite‐derived datasets. Associations between levels of mammal species richness and carbon at the two different spatial resolutions, and for a priori groupings, were assessed via Bayesian two‐stage linear models incorporating quadratic terms.All analyses were conducted in WinBUGS version 1.4.3 through r version 3.3.0 using the package “R2WinBUGS” (Sturtz, Ligges, & Gelman, ); see Appendix S4 for further information on model specification and predictive performance checks. A list of datasets used for analysis is provided in the Data Sources section.RESULTSCamera‐trapping yielded 3,237 independent capture events of 28 species, comprising 24 genera distributed across 16 families. In contrast, our models predicted 30.6 species across the landscape (95% Bayesian Credible Interval, BCI = 28.0–37.0), suggesting that few mammal species were missed by our sampling. The effect of imperfect detection was more pronounced at the camera‐trap level, where predicted richness was consistently greater than observed richness (M = 4.35, range = 0.02–12.26).Spatial concordance between biodiversity and HCS classesHierarchical Bayesian multispecies models indicated reduced mammalian occupancy in the low carbon strata (Figure ). Community (hyper)parameters revealed comparable estimates of mean occupancy between Continuous Logged Forest (M = 0.49, BCI = 0.32–0.63), Dense Forest (0.36, 0.17–0.60) and Developed Land (0.32, 0.12–0.56). However, community occupancy was low in Young Regenerating Forest (0.23, 0.11–0.45) and Oil Palm plantation (0.05, 0.01–0.31).Caterpillar plots of outputs from the hierarchical Bayesian multispecies occupancy model. Graphs show species‐specific baseline occupancy estimates (including 95% Bayesian credible interval) relative to habitat class (a–e). Mean community (hyper)parameter occupancy values and their associated credible intervals are represented in the shaded (orange) background to each plot. Species exhibiting deviations from a baseline occupancy of 0.5 are shown with shaded (blue) bars [Colour figure can be viewed at wileyonlinelibrary.com]Our models demonstrated species‐specific associations with HCS classes (Figure ). For example, occupancy estimates indicate that Sus barbatus Müller (bearded pig) and Macaca nemestrina Linnaeus (southern pig‐tailed macaque) were common in Continuous Logged Forest (S. barbatus: 0.71, 0.53–0.85; M. nemestrina: 0.71, 0.53‐0.86) and Dense Forest (S. barbatus: 0.74, 0.53–0.90; M. nemestrina: 0.74, 0.52–0.92), with occupancy of M. nemestrina also high in Developed Land (0.71, 0.51–0.87). Conversely, species such as Tragulus kanchil Raffles (lesser mouse‐deer: 0.20, 0.08–0.40) and Helarctos malayanus Raffles (sun bear: 0.21, 0.08–0.44) were rare in Dense Forest. In the Oil Palm plantation five species demonstrated low occupancy, four of which were threatened taxa (Figure e). Species‐specific detection summaries for the HCS model are available in Appendix S5.3.Extremes in predicted species richness were identified between the reference habitat classes (Figure ); Continuous Logged Forest was found to have the highest richness (14.12, 13.20–15.07), while Oil Palm plantation supported the most depauperate community (4.54, 3.58–5.52). Estimates of total richness were similar between Dense Forest (11.38, 10.30–12.51) and Developed Land (10.63, 9.52–11.02), while the number of species found in Young Regenerating Forest was significantly lower (8.15, 7.13–9.27). These patterns were consistent across groupings.Boxplots demonstrating species richness in relation to habitat class for: (1) all species; (2) threatened species (IUCN Red Listed as vulnerable, endangered or critically endangered); (3) disturbance‐sensitive species. Letters indicate significant differences between habitat classes within broader species groupings (different letters suggest significance while identical letters indicate non‐significance) [Colour figure can be viewed at wileyonlinelibrary.com]Biodiversity co‐benefits of REDD+The global‐ vs. local‐scale carbon values at camera locations were inconsistent. The 1 km resolution global data tended to produce much higher carbon estimates compared to those derived from higher resolution imagery (global mean = 152.23 t C/ha, range = 50.39–236.53; local mean = 22.95 t C/ha, range = 0.31–94.98). Carbon values from the global‐ and local‐scale maps corresponded broadly with biomass values derived from field inventories (N = 164; rs = 0.55 global; rs = 0.51; local‐scale). However, local‐scale carbon estimates were found to be more precise (RMSE: local = 29.05 t C/ha; global = 130.94 t C/ha). We found no influence of continuous measures of carbon on mammalian occupancy using either global‐ or local‐scale carbon data (see Appendix S5.4/S5.6). Species‐specific detection summaries for the continuous carbon models are available in Appendix S5.5/S5.7.Grain‐dependency between the association of carbon and mammal richness was evident. Using global carbon data, no relationship between the two variables was apparent, regardless of the species grouping (Figure a,c,e). However, at the local‐scale, positive associations with carbon were identified for threatened and disturbance‐sensitive species (Figure d,f). This trend was not consistent across groupings with all species, non‐threatened and disturbance‐tolerant taxa demonstrating no relationship with carbon (Figure b).Bayesian linear model outputs demonstrating significant positive relationships between predicted species richness and carbon stock estimates derived from a 25 m resolution local dataset (d: threatened species; f: disturbance‐sensitive species). All other associations presented were found to be non‐significant. Solid (blue) lines indicate predicted mean posterior distribution values, dashed lines refer to predicted 95% Bayesian credible intervals and vertical grey lines highlight the error associated with each estimated species richness value [Colour figure can be viewed at wileyonlinelibrary.com]DISCUSSIONThe extent to which biodiversity and carbon spatially align is fundamental to our understanding of whether carbon‐based policies can deliver positive results for conservation in human‐modified landscapes. Among the few studies that assess biodiversity and carbon covariance using primary and/or high‐resolution data (Magnago et al., ; Sollmann et al., ), ours is the first to verify an association within a tropical landscape mosaic undergoing certification. We show that the strength, nature and extent of biodiversity co‐benefits are dependent on how carbon stocks are characterised (i.e. categorical or continuous), the spatial resolution of the carbon data employed, and the species considered.Contribution of the HCS approach to biodiversity conservationWhen evaluating community‐level responses to HCS classes, we found comparable levels of mammalian occupancy between Continuous Logged Forest, Developed Land and Dense Forest, while occupancy was reduced in Young Regenerating Forest and Oil Palm. Occupancy can be a viable surrogate for abundance under certain conditions (Efford & Dawson, ). Our results could therefore suggest the persistence of certain mammal species at lower densities within carbon‐poor classes, which confirms previous reports of reduced mammalian abundance in impoverished forest habitats (Bicknell, Struebig, Edwards, & Davies, ).Occupancy and species richness estimates for the total mammal community highlight comparable levels of biodiversity between the Dense Forest and Developed Land classes, supporting previous studies that demonstrate the conservation value of heavily degraded forest for a range of taxonomic groups (Edwards et al., ; Struebig et al., ; Wearn, Carbone, Rowcliffe, Bernard, & Ewers, ). However, we advise caution when interpreting the biodiversity value of Developed Land, which may not be fully realised for long‐lived mammal species until extinction debts, owed to a legacy of disturbance, are repaid (Rosa, Smith, Wearn, Purves, & Ewers, ). The biodiversity value of Developed Land that we found is also crucially dependent on the low levels of hunting at our study site. Hunting has been shown to have substantial impacts on mammal communities elsewhere in the region (Harrison et al., ). Our study adds to the growing body of evidence that shows oil palm plantations to have depauperate mammalian communities, comprised of few generalist species occurring at low densities (Wearn et al., ; Yue, Brodie, Zipkin, & Bernard, ). While our data generally conform to the pattern of declining biodiversity relative to structural complexity, Young Regenerating Forest demonstrated comparably lower biodiversity value regardless of the metric examined. Given the difficulties in differentiating between the Young Regenerating Forest and Scrub strata (see Appendix S2), we believe this finding reflects ambiguities in the HCS classification process.While our analyses demonstrate differences in conservation value between the HCS strata, the ability of land parcels to support biodiversity will also be limited by habitat fragmentation effects. This process is pervasive in human‐modified landscapes, and has contributed to species richness declines of up to 75% (Haddad et al., ). While efforts to account for habitat fragmentation in the HCS prioritisation process are underway, a definitive toolkit is still in development. Therefore, while we have not explicitly accounted for the independent and interactive effects of fragmentation metrics (e.g. patch size, isolation and connectivity) on biodiversity in our analyses, it warrants further consideration as the HCS Approach gains traction across the agricultural sector.Contribution of REDD+ to biodiversity conservationOur results indicate that spatial concordance between biodiversity and carbon can be overlooked if the latter is calculated via low‐resolution data. Using carbon information from a commonly utilised global dataset, no association with mammal diversity was identified, suggesting that REDD+ initiatives would not provide biodiversity co‐benefits in heavily degraded landscapes. However, when high‐resolution carbon maps were employed, a positive relationship with species richness was found for threatened and disturbance‐sensitive taxa, demonstrating the value of REDD+ to those species most vulnerable to land‐use change. When all species were considered, these relationships were obscured by non‐threatened, generalist species that are resilient to disturbance. Our findings provide further support for biodiversity co‐benefits in agricultural land‐use mosaics, as previously demonstrated for a range of taxonomic groups (birds and dung beetles: Gilroy et al., ; amphibians: Basham et al., ), while highlighting important nuances in the carbon–biodiversity relationship. We advocate the use of fine‐grained, field‐validated carbon data when determining the extent and nature of biodiversity co‐benefits and suggest an emphasis on species of conservation concern.Our detailed landscape appraisal is the first to identify biodiversity co‐benefits for mammals, a taxonomic group that occupies key trophic positions in tropical forest ecosystems and is frequently prioritised by conservation. Previous studies have proved less convincing. Across a pantropical network of sites, Beaudrot et al. () found no association between forest carbon and three measures of mammalian diversity. However, by aggregating fine‐scale biomass data at the site level, the authors compromised the resolution of their data, potentially obscuring intra‐site relationships that would be more representative of a REDD+ management unit. Similarly, Sollmann et al. () found little correspondence between above‐ground biomass and mammal occupancy in a certified forest reserve in Malaysian Borneo, despite adopting a comparable methodology to the present study. Contrasting findings may be attributed to spatial variability in hunting pressure.Implications for HCS implementationGiven that RSPO members have little obligation to protect highly disturbed forest of uncertain conservation value, the HCS Approach is a useful tool to designate high carbon, high biodiversity land in areas that would otherwise be converted to plantation. HCS areas can also contribute to national and regional spatial planning initiatives that mitigate the effects of environmental change on tropical biodiversity by promoting connectivity in human‐modified landscapes (Struebig et al., ).Under current HCS guidelines, 62% (8,150 ha) of the remaining forest in the study system would qualify for protection from agricultural conversion, equating to a net gain of 15.72 t C/ha (see Appendix S2) at an annual opportunity cost of US$3.7 million (based on Fisher et al., ). The success of certification depends on financial returns from sustainable production offsetting the economic losses associated with sustainable practices. While the zero deforestation principle of the HCS Approach reduces reputational risk by aligning with consumer goods forum calls to eliminate deforestation from global commodity supply chains, it has been considered economically restrictive for nations with extensive pristine forests (Senior, Brown, Villalpando, & Hill, ), indicating that current guidelines may be too stringent. Strata such as Young Regenerating Forest might, therefore, end up being earmarked for conversion rather than conservation in some circumstances. However, with the conservation value of this stratum likely to increase as forests regenerate, the impact of such a policy change needs to be fully evaluated. Carbon neutral conversion represents an alternative to the current emphasis on zero deforestation. While the specific carbon threshold for delineating forest has proved contentious, Pirker et al. () demonstrated that protecting areas exceeding 100 t C/ha would safeguard 73% of the climatically suitable area for oil palm expansion. Ultimately, compromise begets progression, and while the industry should still strive for zero deforestation, carbon neutral conversion may be more viable in specific countries and circumstances, if agricultural expansion, economic development and forest conservation are to be reconciled.Implications for REDD+ implementationThe considerable enthusiasm for biodiversity co‐benefits often obscures the fact that REDD+ is fundamentally a carbon‐orientated mechanism with limited scope for increasing biodiversity conservation (Venter, Hovani, Bode, & Possingham, ). While we provide further evidence to verify biodiversity co‐benefits in human‐modified landscapes, it is unlikely that REDD+ will be economically viable in carbon‐poor environments. Given current economic pressures and weak carbon markets, REDD+ projects currently prioritise carbon gains at low operating costs. Acting optimally for carbon will therefore place increasing agricultural pressure on secondary or degraded forests that are comparatively low in carbon value but retain appreciable levels of biodiversity (Edwards et al.,). Conservationists must ensure that safeguards are in place to support vulnerable species in disturbed habitats that fall beyond the remit of carbon‐financing mechanisms.The viability of REDD+ in human‐modified landscapes is further hindered by the profitability of oil palm. Under current voluntary markets, avoided deforestation through REDD+ was found to have an opportunity cost of $3,221–8,636 ha−1 over a 30‐year period when compared to potential profits generated from oil palm (Butler et al., ). For REDD+ to be an economically competitive alternative to oil palm cultivation, climate change policies must legitimise REDD+ carbon credits to facilitate their trade on financially lucrative compliance markets (Butler et al., ).CONCLUSIONSOur work highlights the potential for environmental certification and REDD+ financing mechanisms to work in tandem with conservation to mitigate the effects of agricultural expansion on tropical forest carbon stocks and biodiversity. REDD+ is well placed if it prioritises large tracts of contiguous forest, especially if commitments to carbon stock enhancement safeguard degraded forest of biological value. Certification schemes, coupled with land‐use planning tools such as HCS, can help secure sizeable forest patches of high conservation value in agricultural estates, and offer a further safeguard to minimise encroachment. Conservationists should capitalise on both types of carbon‐based policy to maximise the potential for developed lands to provide ecological stepping stones for threatened wildlife between a network of high carbon, high biodiversity areas.ACKNOWLEDGEMENTSThis study was funded by the UK Natural Environment Research Council (NERC) (NE/K016407/1; http://lombok.hmtf.info/) and a NERC EnvEast PhD studentship to N.J.D. G.G.A. is the recipient of a Discovery Early Career Research Award from the Australian Research Council (DE160100904). We thank the Sabah Biodiversity Council, Sabah Forest Department, Yayasan Sabah, Sime Darby and Benta Wawasan for permitting access. We are indebted to Jamiluddin Jami, Arnold James, Mohd. Mustamin, Ampat Siliwong, Sabidee Mohd. Rizan, Najmuddin Jamal and Lois Kinneen for field assistance. We also thank Rob McWilliam, Tri Agus Sugiyanto (both The Forest Trust), Grant Rosoman and Sapta Ananda (both Greenpeace) for training on HCS techniques.AUTHORS' CONTRIBUTIONSN.J.D., M.J.S. and Z.G.D. conceived the ideas and designed the methodology; N.J.D. and E.L.B. collected and processed data, adapting scripts developed by O.R.W.; M.P. and N.J.D. provided carbon data; N.J.D. and G.G.A. undertook analyses; N.J.D., M.J.S. and Z.G.D. led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.DATA ACCESSIBILITYData available from the NERC Environmental Information Data Centre following an embargo period (accessible from: 31/08/2019). Detection data for 28 observed mammal species: https://doi.org/10.5285/62774180-ae72-4873-9482-e8be3935f533 (Deere, Guillera‐Arroita, & Baking, ); HCS stratification of the study site: https://doi.org/10.5285/81cad1ef-b5cc-4592-a71f-204a5d04b700 (Deere et al., ).REFERENCESAvitabile, V., Herold, M., Heuvelink, G. B. M., Lewis, S. L., Phillips, O. L., Asner, G. P., … Willcock, S. (2016). An integrated pan‐tropical biomass map using multiple reference datasets. Global Change Biology, 22, 1406–1420. https://doi.org/10.1111/gcb.2016.22.issue-4Baccini, A., Goetz, S. J., Walker, W. S., Laporte, N. T., Sun, M., Sulla‐Menashe, D., … Houghton, R. A. (2012). Estimated carbon dioxide emissions from tropical deforestation improved by carbon‐density maps. Nature Climate Change, 2, 182–185. https://doi.org/10.1038/nclimate1354Basham, E. W., González del Pliego, P., Acosta‐Galvis, A. R., Woodcock, P., Medina Uribe, C. A., Haugaasen, T, … Edwards, D. P. (2016). Quantifying carbon and amphibian co‐benefits from secondary forest regeneration in the Tropical Andes. Animal Conservation, 19, 548–560. https://doi.org/10.1111/acv.12276Beaudrot, L., Kroetz, K., Alvarez‐Loayza, P., Amaral, L., Breuer, T., Fletcher, C., … Andelman, S. (2016). Limited carbon and biodiversity co‐benefits for tropical forest mammals and birds. Ecological Applications, 26, 1098–1111. https://doi.org/10.1890/15-0935Bicknell, J. E., Struebig, M. J., Edwards, D. P., & Davies, Z. G. (2014). Improved timber harvest techniques maintain biodiversity in tropical forests. Current Biology, 24, R1119–R1120. https://doi.org/10.1016/j.cub.2014.10.067Butler, R. A., Koh, L. P., & Ghazoul, J. (2009). REDD in the red: Palm oil could undermine carbon payment schemes. Conservation Letters, 2, 67–73. https://doi.org/10.1111/conl.2009.2.issue-2Chave, J., Rejou‐Mechain, M., Burquez, A., Chidumayo, E., Colgan, M. S., Delitti, W. B. C., … Vieilledent, G (2014). Improved allometric models to estimate the aboveground biomass of tropical trees. Global Change Biology, 20, 3177–3190. https://doi.org/10.1111/gcb.2014.20.issue-10Deere, N. J., Guillera‐Arroita, G., Baking, E. L., Bernard, H., Pfeifer, M., Reynolds, G., … Struebig, M. J. (2017a). Mammal detection data for the SAFE project site, Sabah, Malaysian Borneo, 2015 [HTMF]. NERC Environmental Information Data Centre. https://doi.org/10.5285/62774180-ae72-4873-9482-e8be3935f533Deere, N. J., Guillera‐Arroita, G., Baking, E. L., Bernard, H., Pfeifer, M., Reynolds, G., … Struebig, M. J. (2017b). High Carbon Stock stratification of the SAFE project site, Sabah, Malaysian Borneo, 2015 [HTMF]. NERC Environmental Information Data Centre. https://doi.org/10.5285/81cad1ef-b5cc-4592-a71f-204a5d04b700Dorazio, R. M., & Royle, J. A. (2005). Estimating size and composition of biological communities by modeling the occurrence of species. Journal of the American Statistical Association, 100, 389–398. https://doi.org/10.1198/016214505000000015Edwards, D. P., Fisher, B., & Wilcove, D. S. (2012). High conservation value or high confusion value? Sustainable agriculture and biodiversity conservation in the tropics. Conservation Letters, 5, 20–27. https://doi.org/10.1111/conl.2012.5.issue-1Edwards, D. P., Magrach, A., Woodcock, P., Ji, Y., Lim, N. T.‐L., Edwards, F. A., … Yu, D. W. (2014). Selective‐logging and oil palm: Multitaxon impacts, biodiversity indicators, and trade‐offs for conservation planning. Ecological Applications, 24, 2029–2049. https://doi.org/10.1890/14-0010.1Efford, M. G., & Dawson, D. K. (2012). Occupancy in continuous habitat. Ecosphere, 3, 1–15.Egoh, B., Reyers, B., Rouget, M., Bode, M., & Richardson, D. M. (2009). Spatial congruence between biodiversity and ecosystem services in South Africa. Biological Conservation, 142, 553–562. https://doi.org/10.1016/j.biocon.2008.11.009Ewers, R. M., Didham, R. K., Fahrig, L., Ferraz, G., Hector, A., Holt, R. D., … Turner, E. C. (2011). A large‐scale forest fragmentation experiment: The stability of altered forest ecosystems project. Philosophical Transactions of the Royal Society B‐Biological Sciences, 366, 3292–3302. https://doi.org/10.1098/rstb.2011.0049Fisher, B., Edwards, D. P., Giam, X. L., & Wilcove, D. S. (2011). The high costs of conserving Southeast Asia's lowland rainforests. Frontiers in Ecology and the Environment, 9, 329–334. https://doi.org/10.1890/100079Fitzherbert, E. B., Struebig, M. J., Morel, A., Danielsen, F., Bruhl, C. A., Donald, P. F., & Phalan, B. (2008). How will oil palm expansion affect biodiversity? Trends in Ecology & Evolution, 23, 538–545. https://doi.org/10.1016/j.tree.2008.06.012Gardner, T. A., Barlow, J., Araujo, I. S., Ávila‐Pires, T. C., Bonaldo, A. B., Costa, J. E., … Peres, C. A. (2008). The cost‐effectiveness of biodiversity surveys in tropical forests. Ecology Letters, 11, 139–150. https://doi.org/10.1111/j.1461-0248.2007.01133.xGardner, T. A., Burgess, N. D., Aguilar‐Amuchastegui, N., Barlow, J., Berenguer, E., Clements, T., … Vieira, I. C. G. T. (2012). A framework for integrating biodiversity concerns into national REDD+ programmes. Biological Conservation, 154, 61–71. https://doi.org/10.1016/j.biocon.2011.11.018Gaveau, D. L. A., Sloan, S., Molidena, E., Yaen, H., Sheil, D., Abram, N. K., … Meijaard, E. (2014). Four decades of forest persistence, clearance and logging on borneo. PLoS ONE, 9, e101654. https://doi.org/10.1371/journal.pone.0101654Gelman, A., & Hill, J. (2007). Data analysis using regression and multilevel/hierarchical models. Cambridge, UK: Cambridge University Press.Gibbs, H. K., Ruesch, A. S., Achard, F., Clayton, M. K., Holmgren, P., Ramankutty, N., & Foley, J. A. (2010). Tropical forests were the primary sources of new agricultural land in the 1980s and 1990s. Proceedings of the National Academy of Sciences of the United States of America, 107, 16732–16737. https://doi.org/10.1073/pnas.0910275107Gillies, C. S., & St Clair, C. C. (2010). Functional responses in habitat selection by tropical birds moving through fragmented forest. Journal of Applied Ecology, 47, 182–190. https://doi.org/10.1111/jpe.2010.47.issue-1Gilroy, J. J., Woodcock, P., Edwards, F. A., Wheeler, C., Baptiste, B. L. G., Medina Uribe, C. A, … Edwards, D. P. (2014). Cheap carbon and biodiversity co‐benefits from forest regeneration in a hotspot of endemism. Nature Climate Change, 4, 503–507. https://doi.org/10.1038/nclimate2200Guillera‐Arroita, G. (2017). Modelling of species distributions, range dynamics and communities under imperfect detection: Advances, challenges and opportunities. Ecography, 40, 281–295. https://doi.org/10.1111/ecog.02445Haddad, N. M., Brudvig, L. A., Clobert, J., Davies, K. F., Gonzalez, A., & Holt, R. D, … Townshend, J. R. (2015). Habitat fragmentation and its lasting impact on Earth's ecosystems. Science Advances, 1, e1500052.Hansen, M. C., Potapov, P. V., Moore, R., Hancher, M., Turubanova, S. A., Tyukavina, A, … Townshend, R. G. (2013). High‐resolution global maps of 21st‐century forest cover change. Science, 342, 850–853. https://doi.org/10.1126/science.1244693Harrison, R. D., Sreekar, R., Brodie, J. F., Brook, S., Luskin, M., O'Kelly, H, … Velho, N. (2016). Impacts of hunting on tropical forests in Southeast Asia. Conservation Biology, 30, 972–981. https://doi.org/10.1111/cobi.12785Kéry, M., & Royle, J. A. (2015). Applied hierarchical modeling in ecology: Analysis of distribution, abundance and species richness in R and BUGS: Volume 1: prelude and static models. London: Academic Press.Kessler, M., Hertel, D., Jungkunst, H. F., Kluge, J., Abrahamczyk, S., Bos, M, … Tscharntke, T. (2012). Can joint carbon and biodiversity management in tropical agroforestry landscapes be optimized? PLoS ONE, 7, e47192. https://doi.org/10.1371/journal.pone.0047192Magnago, L. F. S., Magrach, A., Laurance, W. F., Martins, S. V., Meira‐Neto, J. A. A., Simonelli, M., & Edwards, D. P. (2015). Would protecting tropical forest fragments provide carbon and biodiversity co‐benefits under REDD+? Global Change Biology, 21, 3455–3468. https://doi.org/10.1111/gcb.2015.21.issue-9Mitchard, E. T. A., Feldpausch, T. R., Brienen, R. J. W., Lopez‐Gonzalez, G., Monteagudo, A., Baker, T. R, … Phillips, O. L. (2014). Markedly divergent estimates of Amazon forest carbon density from ground plots and satellites. Global Ecology and Biogeography, 23, 935–946. https://doi.org/10.1111/geb.2014.23.issue-8Murray, J. P., Grenyer, R., Wunder, S., Raes, N., & Jones, J. P. (2015). Spatial patterns of carbon, biodiversity, deforestation threat, and REDD+ projects in Indonesia. Conservation Biology, 29, 1434–1445. https://doi.org/10.1111/cobi.12500Naidoo, R., Balmford, A., Costanza, R., Fisher, B., Green, R. E., Lehner, B. T. R, … Ricketts, T. H. (2008). Global mapping of ecosystem services and conservation priorities. Proceedings of the National Academy of Sciences of the United States of America, 105, 9495–9500. https://doi.org/10.1073/pnas.0707823105Pacifici, K., Zipkin, E. F., Collazo, J. A., Irizarry, J. I., & DeWan, A. (2014). Guidelines for a priori grouping of species in hierarchical community models. Ecology and Evolution, 4, 877–888. https://doi.org/10.1002/ece3.2014.4.issue-7Pfeifer, M., Kor, L., Nilus, R., Turner, E., Cusack, J., Lysenko, I., … Ewers, R. M. (2016b). Mapping the structure of Borneo's tropical forests across a degradation gradient. Remote Sensing of Environment, 176, 84–97. https://doi.org/10.1016/j.rse.2016.01.014Pirker, J., Mosnier, A., Kraxner, F., Havlík, P., & Obersteiner, M. (2016). What are the limits to oil palm expansion? Global Environmental Change, 40, 73–81. https://doi.org/10.1016/j.gloenvcha.2016.06.007Potter, L. (2015). Managing oil palm landscapes: A seven‐country survey of the modern palm oil industry in Southeast Asia, Latin America and West Africa. CIFOR.Rodríguez‐Castañeda, G., Hof, A. R., Jansson, R., & Harding, L. E. (2012). Predicting the fate of biodiversity using species' distribution models: Enhancing model comparability and repeatability. PLoS ONE, 7, e44402. https://doi.org/10.1371/journal.pone.0044402Rosa, I. M., Smith, M. J., Wearn, O. R., Purves, D., & Ewers, R. M. (2016). The environmental legacy of modern tropical deforestation. Current Biology, 26, 2161–2166. https://doi.org/10.1016/j.cub.2016.06.013Rosoman, G., Sheun, S. S., Opal, C., Anderson, P., & Trapshah, R. (Eds.) (2017). The HCS approach toolkit. Singapore: HCS Steering Group.Royle, J., & Dorazio, R. (2008). Hierarchical modelling and inference in ecology. New York, NY: Academic Press.Royle, J. A., Dorazio, R. M., & Link, W. A. (2007). Analysis of multinomial models with unknown index using data augmentation. Journal of Computational and Graphical Statistics, 16, 67–85. https://doi.org/10.1198/106186007X181425RSPO (2017). Certification in numbers. Retrieved from www.rspo.org/about/impacts.Ruiz‐Jaen, M. C., & Potvin, C. (2010). Tree diversity explains variation in ecosystem function in a neotropical forest in Panama. Biotropica, 42, 638–646. https://doi.org/10.1111/j.1744-7429.2010.00631.xSenior, M. J., Brown, E., Villalpando, P., & Hill, J. K. (2015). Increasing the scientific evidence base in the “high conservation value”(HCV) approach for biodiversity conservation in managed tropical landscapes. Conservation Letters, 8, 361–367. https://doi.org/10.1111/conl.12148Sollmann, R., Mohamed, A., Niedballa, J., Bender, J., Ambu, L., Lagan, P, … Wilting, A. (2017). Quantifying mammal biodiversity co‐benefits in certified tropical forests. Diversity and Distributions, 23, 317–328. https://doi.org/10.1111/ddi.2017.23.issue-3Strassburg, B. B. N., Kelly, A., Balmford, A., Davies, R. G., Gibbs, H. K., Lovett, A, … Rodrigues, A. S. L. (2010). Global congruence of carbon storage and biodiversity in terrestrial ecosystems. Conservation Letters, 3, 98–105. https://doi.org/10.1111/conl.2010.3.issue-2Struebig, M. J., Kingston, T., Petit, E. J., Le Comber, S. C., Zubaid, A., Mohd‐Adnan, A., & Rossiter, S. J. (2011). Parallel declines in species and genetic diversity in tropical forest fragments. Ecology Letters, 14, 582–590. https://doi.org/10.1111/j.1461-0248.2011.01623.xStruebig, M. J., Turner, A., Giles, E., Lasmana, F., Tollington, S., Bernard, H., & Bell, D. (2013). Quantifying the biodiversity value of repeatedly logged rainforests: Gradient and comparative approaches from Borneo. Advances in Ecological Research, 48, 183–224. https://doi.org/10.1016/B978-0-12-417199-2.00003-3Struebig, M. J., Wilting, A., Gaveau, D. L., Meijaard, E., Smith, R. J., Fischer, M., … Wong, A., (2015). Targeted conservation to safeguard a biodiversity hotspot from climate and land‐cover change. Current Biology, 25, 372–378. https://doi.org/10.1016/j.cub.2014.11.067Sturtz, S., Ligges, U., & Gelman, A. (2005). R2WinBUGS: a package for running WinBUGS from R. Journal of Statistical software, 12, 1–16.Venter, O., Hovani, L., Bode, M., & Possingham, H. (2013). Acting optimally for biodiversity in a world obsessed with REDD+. Conservation Letters, 6, 410–417. https://doi.org/10.1111/conl.2013.6.issue-6Venter, O., & Koh, L. P. (2012). Reducing emissions from deforestation and forest degradation (REDD+): Game changer or just another quick fix? Annals of the New York Academy of Sciences, 1249, 137–150. https://doi.org/10.1111/j.1749-6632.2011.06306.xWearn, O. R., Carbone, C., Rowcliffe, J. M., Bernard, H., & Ewers, R. M. (2016). Grain‐dependent responses of mammalian diversity to land‐use and the implications for conservation set‐aside. Ecological Applications, 26, 1409–1420. https://doi.org/10.1890/15-1363Wilcove, D. S., Giam, X., Edwards, D. P., Fisher, B., & Koh, L. P. (2013). Navjot's nightmare revisited: Logging, agriculture, and biodiversity in Southeast Asia. Trends in Ecology & Evolution, 28, 531–540. https://doi.org/10.1016/j.tree.2013.04.005Wilson, K. A., Meijaard, E., Drummond, S., Grantham, H. S., Boitani, L., Catullo, G, … Watts, M. (2010). Conserving biodiversity in production landscapes. Ecological Applications, 20, 1721–1732. https://doi.org/10.1890/09-1051.1Wulder, M. A., Franklin, S. E., White, J. C., Cranny, M. M., & Dechka, J. A. (2004). Inclusion of topographic variables in an unsupervised classification of satellite imagery. Canadian Journal of Remote Sensing, 30, 137–149. https://doi.org/10.5589/m03-063Yaap, B., Struebig, M. J., Paoli, G., & Koh, L. P. (2010). Mitigating the biodiversity impacts of oil palm development. CAB Reviews, 5, 1–11.Yue, S., Brodie, J. F., Zipkin, E. F., & Bernard, H. (2015). Oil palm plantations fail to support mammal diversity. Ecological Applications, 25, 2285–2292. https://doi.org/10.1890/14-1928.1Zipkin, E. F., Royle, J. A., Dawson, D. K., & Bates, S. (2010). Multi‐species occurrence models to evaluate the effects of conservation and management actions. Biological Conservation, 143, 479–484. https://doi.org/10.1016/j.biocon.2009.11.016DATA SOURCESAvitabile, V., Herold, M., Heuvelink, G. B. M., Lewis, S. L., Phillips, O. L., Asner, G. P., … Willcock, S. (2016). An integrated pan‐tropical biomass map using multiple reference datasets. Global Change Biology, 22, 1406–1420. https://doi.org/10.1111/gcb.2016.22.issue-4Pfeifer, M., Kor, L., Nilus, R., Turner, E., Cusack, J., Lysenko, I., … Ewers, R. M. (2016a). Mapping the structure of Borneo's tropical forests across a degradation gradient. Remote Sensing of Environment, 176, 84–97. https://doi.org/10.1016/j.rse.2016.01.014

Journal

Journal of Applied EcologyWiley

Published: Jan 1, 2018

Keywords: ; ; ; ; ; ; ; ; ;

There are no references for this article.