The authors have declared that no competing interests exist.
The rapid expansion of oil palm cultivation in the Neotropics has generated great debate around possible biodiversity impacts. Colombia, for example, is the largest producer of oil palm in the Americas, but the effects of oil palm cultivation on native fauna are poorly understood. Here, we compared how richness, abundance and composition of terrestrial mammal species differ between oil palm plantations and riparian forest in the Colombian Llanos region. Further, we determined the relationships and influence of landscape and habitat level variables on those metrics. We found that species richness and composition differed significantly between riparian forest and oil palm, with site level richness inside oil palm plantations 47% lower, on average, than in riparian forest. Within plantations, mammalian species richness was strongly negatively correlated with cattle abundance, and positively correlated with the density of undergrowth vegetation. Forest structure characteristics appeared to have weak and similar effects on determining mammal species richness and composition along riparian forest strips. Composition at the landscape level was significantly influenced by cover type, percentage of remaining forest and the distance to the nearest town, whereas within oil palm sites, understory vegetation, cattle relative abundance, and canopy cover had significant effects on community composition. Species specific abundance responses varied between land cover types, with oil palm having positive effects on mesopredators, insectivores and grazers. Our findings suggest that increasing habitat complexity, avoiding cattle and retaining native riparian forest–regardless of its structure–inside oil palm-dominated landscapes would help support higher native mammal richness and abundance at both local and landscape scales.
Habitat loss caused by agricultural expansion is one of the main drivers of global biodiversity loss [
In the Neotropics (Latin America), oil palm production is rapidly expanding [
The Llanos Orientales region (eastern plains) of Colombia is renowned for its species and ecosystem diversity, comprising large areas of savannas, grasslands, wetlands and riparian forest (known locally as gallery forest) [
Mammals are a good indicator of ecosystem quality or change, given their diversity and the complexity of ecological niches they occupy [
In this study, we use an extended camera trapping survey in the western Llanos Orientales- Colombia’s leading oil palm production region, to compare species richness, abundance and composition of terrestrial mammals between oil palm plantations and riparian forests strips, the two most dominant land cover types in the region. Further, we determine the main landscape and habitat correlates driving mammalian species richness, abundance and composition within and between these land cover types in an attempt to identify management practices that may help minimize the impact of this expanding agricultural practice.
We conducted this study across a ~2,000 km2 area in the rural areas surrounding the towns of Restrepo, Cumaral, Cabuyaro, Acacias, Castilla la Nueva, and San Carlos de Guaroa, in the Department of Meta, situated in the eastern plains or Llanos Orientales region of Colombia, (
We sampled 33 sites in oil palm plantations (hereafter referred to as plantations) and 23 in riparian forests (hereafter referred to as forests) with sampling effort proportionate to the spatial extent of these land covers within the study area (
We used camera traps to detect medium and large (> 1kg) terrestrial mammals in the dry/transition seasons (~Sept-Mar) across the period September 2014 to January 2016. Since we had a limited number of cameras, the study area was not sampled simultaneously across all sites, and surveys were organized sequentially in different sessions. We used seven cameras to sample each site, as determined by a pilot study that we conducted in the study area (Pardo et al. not publ.). This sampling intensity was implemented to ensure greater sampling completeness compared with traditional practice of using a single camera per site [
Cameras (Reconyx HC500 HyperfireTM, United States [US]) were active for a minimum of 30 days at each site and were configured according to the following criteria: high sensitivity, one-second intervals between consecutive photographs (3 per trigger), no delay or quiet period between triggers, a minimum distance of 1.5 m from an animal’s potential path, and a height of 25‒30 cm depending on the terrain. All cameras were fixed to trees or wooden poles (in the case of cameras inside plantations) with a steel security cable (PythonTM, US). Arboreal and other species not likely detected by camera trap were recorded opportunistically by direct observations, but were not use for analysis.
This research was conducted in compliance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, 7th Edition, 2004 and the Qld Animal Care and Protection Act, 2001. This study received animal ethics approval from Animal Ethics Committee of James Cook University. The owners of the land and the oil palm companies at each site gave permission to conduct the study on their lands. No specific permissions were required for these locations/activities. The field studies did not involve handling/manipulation of endangered or protected species.
To estimate species richness, we first computed species accumulation curves using EstimateS [
We used capture frequencies of individual species as a proxy for a relative abundance index (e.g. [
We selected five landscape covariates previously shown to influence mammal richness and composition: 1) percentage of forest [
To quantify the percentage of forest at each site, we created a 500 m-radius buffer around each camera within each individual transect and then merged the buffers into one single area for analysis. Distance to road and towns was calculated as the average Euclidean distance (m) to the nearest road or town (respectively) for all cameras within the site. These measurements were all obtained using Quantum GIS 2.0.1 [
We used different sets of habitat covariates for assessing patterns within plantations and riparian forests. Within plantations habitat variables were related to crop management practices. These included: 1) the presence of undergrowth vegetation (a factor with 2 levels–see
We evaluated the influence of landscape and habitat covariates on mean mammalian species richness using individual Poisson generalized linear mixed models (GLMMs) with habitat and landscape variables as fixed factors, and site as random factor. Prior to model generation, we checked for correlated predictor variables following the protocol of Zuur [
We generated models with all valid combinations of the covariates without interaction effects (32 models for landscape covariates, and 16 models for habitat covariates in the 2 land-cover types) and used an information-theoretic approach to determine the most parsimonious model based on Akaike’s information criterion, corrected for small sample size (AICc) [
Analyses of abundance and composition were undertaken in two separate procedures. First, we used a Non-metric Multidimensional Scaling (NMDS) ordination based on Bray Curtis dissimilarity matrix among sites to visualize overall differences in the structure and composition of the assemblage between oil palm and forest (i.e. the distribution of capture records across species and sites). The NMDS is a flexible technique that uses rank orders to evaluate dissimilarities between different communities instead of absolute distances [
Second, to test for the effect of landscape and habitat covariates on overall community composition and on individual species relative abundances, we used a multivariate version of generalized linear modeling (GLM) via
We sampled a total of 12,403 camera days and identified 24 ground dwelling species (23 medium to large sized and one small mouse) and two arboreal monkeys, representing seven taxonomic orders and 16 families (
The sampling completeness for mammals in the study area was relatively high, suggesting that the sampling intensity within sites as well as the number of sites captured most of the total species expected in the region (mean = 84%; SD = 15.97). Rarefaction curves showed a representative sample effort with clear asymptotes. Associated confidence intervals of these curves did not overlap, indicating that total richness between plantations and forest was significantly different (
The detection frequency (i.e. relative abundance) of the majority of species was low across the study area (
Bars indicate the upper standard deviation range. Blue lines separate taxonomic orders (from left to right): Pilosa, Cingulata, Carnivora, Artiodactyla, Rodentia, Marsupialia. Note: A mouse species and two species of primates were also detected, but they are not included in this figure because the first could not be identified by camera trap and the latter are not ground dwelling mammals.
No single model offered the best explanation for species richness at the landscape level. Results from the averaged model (using 14 suitable candidate models of Δ AICc <7) revealed that land-cover type was clearly the main driver of differences in species richness (∑
Relationship between mammalian species richness as a function of landscape covariates in Llanos, Colombia: a) percentage of forest, b) NDVI, c) distance to towns, and d) distance to roads, according to land-cover type (oil palm plantations vs forest). The trend lines are predicted values of the GLMM model averaged (holding other covariates constant) and shaded areas represent the 95% confidence intervals. Dotted points represent the actual values of the covariate. Effect of land-cover type is strong, while the slope of continues variables does not show an important effect on species richness.
Estimates correspond to the conditional averaged parameter coefficient and relative importance is based on the
Estimate | Adjusted SE | ||
---|---|---|---|
Intercept | 2.21 | 0.08 | |
Land-cover type (oil palm plantation) | ‒0.74 | 0.13 | 1.00 |
NDVI | 0.07 | 0.07 | 0.36 |
Distance to nearest town (km) | 0.03 | 0.06 | 0.24 |
Forest (%) | 0.02 | 0.08 | 0.23 |
Dist. road (km) | ‒0.01 | 0.06 | 0.23 |
Intercept | 1.27 | 0.20 | |
Cattle detection frequency | ‒0.27 | 0.15 | 0.69 |
Understory vegetation (medium-high) | 0.41 | 0.24 | 0.55 |
Height (m) | ‒0.13 | 0.10 | 0.39 |
Distance to nearest patch (km) | ‒0.11 | 0.10 | 0.35 |
Canopy cover (%) | 0.08 | 0.13 | 0.26 |
Intercept | 2.22 | 0.07 | |
Number of trees | ‒0.05 | 0.08 | 0.22 |
DBH (cm) | 0.05 | 0.08 | 0.22 |
Canopy cover (%) | ‒0.02 | 0.08 | 0.18 |
Height (m) | ‒0.01 | 0.09 | 0.18 |
Similar to the landscape covariates, no single model was identified as demonstrably better than any other, and 15 candidate models were retained based on ΔAICc values (
No single model emerged as a possible driver of mammalian richness inside forest (
Overall, ordination analysis indicated important dissimilarities in composition between plantations and forest, with plantation sites relatively more scattered and separated from each other (i.e. more different in composition) compared to forest sites (
Plot is based on capture frequencies of species using Bray-Curtis non-metric multidimentional analysis (NMDS) (stress = 0.22). Polygons connect the vertices of each cover type and ellipses emphasize the centroids of the community in each land cover. Species outside the boundaries were very rare in the landscape. Codes correspond to the initial letters of the scientific names of each species (refer to
Individual species´ abundance response varied between species. For example, the strongest negative effect of plantations on species abundances were shown by agouti and paca (
Values indicate GLM model coefficients and colors represent the effect sizes on the relative abundance of each species (univariate analysis). Abbreviations: Plantations = oil palm plantation, one of the 2 levels of the categorical variable “cover type” (i.e., riparian forest, and oil palm plantations); x.for = percentage of forests in the 500 m-radius buffer; Dist.road and Dist.town = the average nearest distance to roads and towns (respectively); NDVI: Normalized Difference Vegetation Index. Variables were standardized for direct comparison.
Understory vegetation, relative abundance of cattle and canopy cover had a significant effect on the community composition (i.e. the overall combined effects of each factor, simultaneously assessed across all species) of mammals within plantations (Deviance [Dev] = 20.88, Pr (>Dev) = 0.045; Deviance [Dev] = 29.57, Pr (>Dev) = 0.002; Deviance [Dev] = 19.87, Pr (>Dev) = 0.043, respectively). Most of the individual species’ coefficients were close to zero, with high standard errors, likely due to the low detections. Therefore, inferences for individual responses inside oil palm should be taken cautiously. However, the detection of rare species, (i.e those with less than three records) such as grison, puma, red-brocket deer, peccary, paca (hence not suitable for statistical analysis), and capybara inside plantations were restricted to sites near to forest (i.e. mean distance among species of 430 m–SD 169 m).
There was no significant evidence of any habitat variable influencing mammalian community composition inside the forests (DBH = [Dev] = 12.33, Pr (>Dev) = 0.59; tree abundance = [Dev] = 0, Pr (>Dev) = 0.95; canopy cover = [Dev] = 10.44, Pr (>Dev) = 0.75 and tree height = [Dev] = 22.43, Pr (>Dev) = 0.18).
This study aimed to understand the structure and responses of mammal assemblages to an oil palm dominated landscape in Colombia. Our results indicate that richness and composition are significantly reduced in oil palm plantations compared to adjacent riparian forest. However, responses of individual species varied, with the relative abundances of most species responding negatively to oil palm, while other species appeared unaffected or even displayed positive responses to oil palm occurrence, such as mesopredators, insectivores and a grazer. This reduction in diversity is similar to results previously reported for parts of Southeast Asia [
The relative little differences in total richness (at the landscape level in particular) between plantations and forest, may be consequence of a long history of landscape transformation, specifically for pasture creation [
Our study highlights the importance of secondary forest presence in human-dominated landscapes for biodiversity conservation. In this study, not only were forests important for mammal species richness, but also for their relative abundance. The lack of evidence found for any particular driver explaining richness within riparian forest (see also [
Contrary to our predictions, the percentage of forest cover in landscape failed to strongly explain species richness at the landscape level. Although as with NDVI, we found evidence of a positive relationship. This finding corresponds to previous studies (e.g. [
We identified two factors explaining richness and composition inside oil palm that can be useful for improving management practices to help to sustain mammal diversity (in terms of both richness and composition): reducing cattle grazing pressure, and maintaining a medium-to-high density of undergrowth vegetation. These factors are clearly linked, as cattle reduces undergrowth in plantations through grazing and soil compaction (LEP pers.obs). Grazing has also previously been found to negatively affect the taxonomic and functional diversity of small mammals in Argentina [
Another potential management practice highlighted by our results was the level of canopy cover within plantation, which had a significant effect on the composition but not in richness. This result implies that plantations with higher levels of canopy cover may be used as a mechanism to increase species movement into and across oil palm plantations. Further study is required to fully examine the effects of canopy cover. One possibility would be to examine assemblage structure between oil palm varieties and/or hybrids that differ in leaf size.
The clear effect of understory vegetation for improving species richness and the potential of manipulating canopy cover inside plantation to promote abundance of species, would support approaches such as “wildlife friendly” production (e.g. [
We assessed individual mammal species responses to potential land-cover transformation (forests to oil palm plantations) as a proxy to evaluate their tolerance or sensitivity to oil palm. The giant anteater, for example, was widely detected in both land cover types across the landscape, and even showed a slightly positive response in abundance to plantations. This positive response confirms that the species can utilize other land use types, such as forest, plantations or pastures [
Our results for giant anteater contrast those from Mendes-Oliveira et al. [
Mesopredators in general showed tolerance to plantations. This result is similar to those of Mendes-Oliveira et al. [
The most likely mechanism driving the aparent high relative abundance of mesopredators could be related to sufficient availability of resources through bottom-up effects [
Overall, numerous terrestrial mammal species were found in both plantations and riparian forests, probably due to the historical context of land use in the study area (apart from likely hunting pressures), which could have limited present-day community to generalist and more ecologically flexible species [
Certification schemes for sustainable agriculture (e.g., the Roundtable on Sustainable Palm Oil—RSPO;
On the other hand, we only detected one species of conservation concern, the giant anteater, which was frequently detected in oil palm landscapes. Therefore, a question that remains is, which species should be prioritized? Most species were not categorized as conservation concern, which under present certification schemes may be considered irrelevant for conservation (see [
If oil palm is mainly replacing pastures and other crops in Colombia [
In terms of the richness and composition of mammal species, we found that the areas of San Carlos de Guaroa and Cabuyaro warrant special attention. In these zones, we detected rare and ecologically important species, such as the puma (see [
Oil palm development provides social benefits in Colombia, and plays an important role as source of employment [
This study provides the first comprehensive analysis of the landscape- and habitat-level effects of oil palm cultivation on terrestrial mammals in Colombia. We found that oil palm plantations supported significantly fewer mammal species and different composition than riparian forests. However, we identified that some species, particularly mesopredators, anteaters, and deer were relatively common in oil palm plantations. We found that secondary riparian forests have a fundamental role in mammal conservation in this landscape, regardless of its structure or area. Therefore, if oil palm expansion occurs at the expense of remnant riparian vegetation there will be drastic deleterious consequence for mammal species in the Llanos region. Based on our results, we recommend that to maintain and increase native mammal diversity inside the plantations, oil palm growers should promote undergrowth vegetation and avoid cattle presence inside plantations, along with respecting designated buffer areas that allow for the conservation and restoration of riparian forests. The present-day assemblage in the study area was limited to relative resilient species. In the absence of pristine or highly threatened species, we suggest the development of new ways of recognition for implementation of good practices that could promote the conservation value and awareness of degraded landscapes.
a) Aerial photographs (August 2014) of the landscape highlighting riparian forest and oil palm plantations structure. b) Differences in management schemes of understory vegetation in oil palm plantations in Llanos, Colombia. Photo credit: L.E.Pardo.
(PDF)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
(DOCX)
Coefficients are from the saturated model using the multispecies generalized linear modelling prior to shrinkage with Lasso penalty (R package mvabund). SE is the standard error of the coefficient. For scientific names and details of the species, refer to
(DOCX)
We are grateful to all companies for their valuable support and assistance during field work and for allowing us to work and stay on their lands, to the National Federation of Oil Palm Growers (FEDEPALMA) and the project “Biodiversity Conservation in the Areas of Oil Palm Plantations” (GEF/BID PPB). We also thank the field guides and research assistants, especially Angela Rojas-Rojas and Juan Albarracin. We are grateful to Mike Meredith, Diego Zarrate, Nicolás Youñes, and David Warton for their comments. The lead author also thanks the Administrative Department of Science, Technology and Innovation–COLCIENCIAS, Colombia, for funding his PhD studies.