Contributed Paper
Mitigating the impact of oil-palm monoculture
on freshwater fishes in Southeast Asia
Xingli Giam,∗ † ‡‡§§ Renny K. Hadiaty,‡ Heok Hui Tan,† Lynne R. Parenti,§ Daisy Wowor,‡
Sopian Sauri,‡ Kwek Yan Chong,∗∗ Darren C. J. Yeo,∗∗ and David S. Wilcove∗ †† §§
∗
Department of Ecology and Evolutionary Biology, Princeton University, Princeton, NJ 08544, U.S.A.
†Lee Kong Chian Natural History Museum, National University of Singapore, S117546, Singapore
‡Museum Zoologicum Bogoriense, Research Center for Biology, Indonesian Institute of Sciences, Cibinong 16911, Indonesia
§Division of Fishes, Department of Vertebrate Zoology, National Museum of Natural History, P.O. Box 37012, MRC 159, Smithsonian
Institution, Washington, D.C. 20013, U.S.A.
∗∗
Department of Biological Sciences, National University of Singapore, S117543, Singapore
††Woodrow Wilson School of Public and International Affairs, Princeton University, Princeton, NJ 08544, U.S.A.
Abstract: Anthropogenic land-cover change is driving biodiversity loss worldwide. At the epicenter of this
crisis lies Southeast Asia, where biodiversity-rich forests are being converted to oil-palm monocultures. As
demand for palm oil increases, there is an urgent need to find strategies that maintain biodiversity in
plantations. Previous studies found that retaining forest patches within plantations benefited some terrestrial
taxa but not others. However, no study has focused on aquatic taxa such as fishes, despite their importance
to human well-being. We assessed the efficacy of forested riparian reserves in conserving freshwater fish biodiversity in oil-palm monoculture by sampling stream fish communities in an oil-palm plantation in Central
Kalimantan, Indonesia. Forested riparian reserves maintained preconversion local fish species richness and
functional diversity. In contrast, local and total species richness, biomass, and functional diversity declined
markedly in streams without riparian reserves. Mechanistically, riparian reserves appeared to increase local
species richness by increasing leaf litter cover and maintaining coarse substrate. The loss of fishes specializing
in leaf litter and coarse substrate decreased functional diversity and altered community composition in oilpalm plantation streams that lacked riparian reserves. Thus, a land-sharing strategy that incorporates the
retention of forested riparian reserves may maintain the ecological integrity of fish communities in oil-palm
plantations. We urge policy makers and growers to make retention of riparian reserves in oil-palm plantations
standard practice, and we encourage palm-oil purchasers to source only palm oil from plantations that employ
this practice.
Keywords: agriculture, buffers, fishes, functional diversity, land sharing, riparian reserves, stream, structural
equation model
Mitigación del Impacto del Monocultivo de Palma de Aceite sobre los Peces de Agua Dulce en el Sureste Asiático
Resumen: El cambio en la cobertura antropogénica del suelo está causando una pérdida mundial de
biodiversidad. En el epicentro de la crisis se encuentra el sureste asiático, en donde los bosques ricos en
biodiversidad se convierten a monocultivos de palma de aceite. Junto con el incremento de la demanda por
la palma de aceite hay una necesidad urgente para encontrar estrategias que mantengan la biodiversidad
dentro las plantaciones. Estudios previos encontraron que la retención de fragmentos de bosque dentro de
las plantaciones benefició a algunos taxones terrestres pero no a los demás. Sin embargo, ningún estudio se
ha enfocado en los taxones acuáticos, como los peces, a pesar de su importancia para el bienestar humano.
Evaluamos la eficacia de las reservas ribereñas de bosque en la conservación de la diversidad de peces de agua
dulce en los monocultivos de palma de aceite al muestrear a las comunidades de peces en una plantación
‡‡Current address: School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA 98105, U.S.A.
§§Address correspondence to: Xingli Giam, email giamxingli@gmail.com or David S. Wilcove, email dwilcove@princeton.edu
Paper submitted August 14, 2014; revised manuscript accepted December 14, 2014.
1357
Conservation Biology, Volume 29, No. 5, 1357–1367
C 2015 Society for Conservation Biology.
DOI: 10.1111/cobi.12483
1358
Conserving Fishes in Oil-Palm Monocultures
de palma de aceite en Kalimantan Central, Indonesia. Las reservas ribereñas de bosque mantuvieron la
diversidad funcional y la riqueza local de especies de peces previa a la conversión. En contraste, la riqueza
total y local de especies, la biomasa y la diversidad funcional declinaron marcadamente en los arroyos sin
reservas ribereñas. Mecánicamente, pareció que las reservas ribereñas incrementaron la riqueza local de
especies al incrementar la cobertura de hojas caı́das y mantener el sustrato áspero. La pérdida de peces
especializados en la cobertura de hojas caı́das y el sustrato áspero disminuyó la diversidad funcional y alteró
la composición comunitaria en los arroyos de plantaciones de palma de aceite que carecı́an de reservas
ribereñas. Ası́, una estrategia de tierras compartidas que incorpora la retención de las reservas ribereñas de
bosque puede mantener la integridad ecológica de las comunidades de peces en las plantaciones de palma
de aceite. Instamos a quienes hacen las polı́ticas y a los cultivadores a que hagan de la retención de reservas
ribereñas en las plantaciones de palma de aceite una práctica estandarizada, y alentamos a los compradores
del aceite de palma a que lo hagan sólo de las plantaciones que empleen esta práctica.
Palabras Clave: agricultura, amortiguadores, arroyo, diversidad funcional, modelo de la ecuación estructural,
peces, reservas, ribereño, tierras compartidas
Introduction
Anthropogenic land-cover change is the major driver
of biodiversity loss worldwide (Millennium Ecosystem
Assessment 2005). The destruction of tropical forests
deserves particular attention for at least 3 reasons:
tropical forests hold the majority of the world’s species
(Wilson 1988); biodiversity declines markedly when
tropical forests are converted to other cover types (e.g.,
cropland) (Gibson et al. 2011); and tropical deforestation
continues unabated due to increasing demands for food,
timber, and other commodities (Balmford et al. 2012;
Putz & Romero 2014). One prominent approach to
balancing biodiversity conservation with commodity
production is to identify strategies that reduce biodiversity losses in production landscapes. Understanding
the efficacy of such strategies is imperative for making
informed decisions regarding managing production
landscapes for conservation (land-sharing approach)
versus managing them to maximize commodity yields
while preserving intact forests outside production lands
(land-sparing approach) (Phalan et al. 2011).
The need to quantify the efficacy of measures to protect biodiversity in working landscapes is especially acute
in Southeast Asia, where both species richness and deforestation rates are among the highest globally (Sodhi
et al. 2004). A major threat to Southeast Asia’s biodiversity
is the conversion of forests (mostly selectively logged) to
monoculture crops, especially oil-palm (Elaeis guineensis) (Wilcove et al. 2013). Conversion of forests to oilpalm monoculture reduces species richness and changes
the composition of both terrestrial and aquatic communities (e.g., Peh et al. 2006; Fayle et al. 2010; Mercer
et al. 2014). Thus far, measures to maintain biodiversity
within oil-palm plantations (e.g., by leaving patches of
forest) have been ineffective for birds (Edwards et al.
2010) and butterflies (Koh 2008), effective for riparian
ants (Gray et al. 2015), and partially effective for riparian
dung beetles (Gray et al. 2014). Taken together, these
results suggest that the land-sparing versus land-sharing
Conservation Biology
Volume 29, No. 5, 2015
debate remains unresolved. However, no one has assessed the efficacy of a land-sharing strategy to mitigate
the impact of oil-palm monoculture on freshwater biota,
despite the fact that Southeast Asia is a global hotspot
for freshwater fish biodiversity (WWF & TNC 2013) and
that many of its people depend on freshwater fishes for
protein (MacKinnon et al. 1997). If oil-palm agriculture
indeed threatens fish species and if these impacts are not
mitigated, consequences could be dire for both biodiversity and humans.
The most common strategy used to protect water quality in production landscapes is retaining forests along
streams (Allan & Flecker 1993; Naiman & Décamps 1997).
Forested riparian reserves (hereafter, riparian reserves)
reduce sedimentation, chemical runoff, and maintain
temperature regimes in streams within temperate production landscapes (Naiman & Décamps 1997; Sweeney
& Newbold 2014); their efficacy in the tropics is less
well-studied. RR also maintain terrestrial subsidies to the
freshwater system (e.g., inputs of leaf litter, wood, and
terrestrial insects) (Dudgeon 2000). However, the degree
to which riparian reserves protect aquatic biodiversity
seems to vary depending on geography and land-use
context. For example, riparian reserves effectively protect fishes in Brazilian (Casatti et al. 2012) and Costa
Rican (Lorion & Kennedy 2009) pastures, but do not protect fishes in pine monoculture in New Zealand (Rowe
et al. 2002). More studies are needed, especially from the
beleaguered forests of Southeast Asia.
In Indonesia and Malaysia—the world’s 2 largest palmoil-producing countries—retention of riparian reserves is
required by law (Government of Malaysia 1965; Republic
of Indonesia 1990). Riparian reserves are also mandatory
for growers who join the Roundtable for Sustainable
Palm Oil (RSPO) certification program (www.rspo.org),
a voluntary program whereby growers agree to employ best management practices aimed at improving
environmental and social outcomes in exchange for
recognition of the environmental sustainability of their
products. But whether riparian reserves actually conserve
Giam et al.
freshwater biota in tropical oil-palm monocultures is
unknown.
We attempted to bridge this crucial knowledge gap
by asking whether freshwater fish biodiversity can be
conserved by retaining forested riparian reserves in oilpalm plantations. We compared freshwater fish species
richness, community composition, functional diversity,
and biomass in streams located in riparian reserves
against streams without riparian reserves in an oil-palm
plantation in Central Kalimantan, Indonesia. By comparing our contemporary field data with historical data, we
examined whether riparian reserves effectively retained
pre-conversion species richness and functional diversity.
We also examined causal ecological mechanisms through
which riparian reserves affect species richness via the
instream environment.
Methods
Study Landscape
We conducted research in a 14,445-ha oil-palm dominated landscape in Central Kalimantan Province, Indonesia, on the island of Borneo. The oil-palm plantation
(PT Unggul Lestari; 10,838 ha) was established in 2007
on lands previously covered by logged and secondary
forests, community forest gardens, and shrublands.
Forested conservation areas and community-owned lands
(3607 ha) are embedded within the plantation.
The landscape is drained by tributaries of the Mentaya
River, which empties into the Java Sea. Riparian reserves
comprise selectively logged and secondary forests that
have been retained along tributaries. Riparian reserves
(>50 m wide) were present along almost all large tributaries (>15 m wide) in the estate. For small tributaries
(3–10 m wide), however, the riparian reserves were irregular in shape and width (from 10 to >100 m wide)
and were present along some stream lengths but not
others, thereby providing an opportunity to test their
efficacy. Where riparian reserves were not retained, oilpalms were planted to the stream edge. Pesticides and
herbicides were not applied within 50 m of any stream
(including streams lacking riparian reserves), which allowed a fair comparison of biodiversity values across
sampling sites.
Site Description and Fish Sampling
Fieldwork was conducted in June 2013. We established
12, 100 m long sampling sites in small streams (3.5–8 m
wide, 1 m deep) within the plantation (Fig. 1). Eight
sites had forested riparian reserves (RR sites) and 4 sites
lacked riparian reserves (i.e., streams surrounded by oil
palms [OP sites]). The 8 RR sites were further classified
into 2 habitat types: logged forest (RRLF ) (3 sites) and
1359
community forest (RRCF ) (5 sites). Both were dominated
by selectively logged or secondary forest, but RRCF sites
were continuously disturbed by local people through cultivation of rubber (Hevea brasiliensis), tapioca (Manihot
esculenta), banana (Musa cultivars), rattan, or subsistence logging. Each sampling site was in an independent
tributary to ensure spatial independence of data.
We sampled fishes on clear-weather days using 3
capture methods, performed in the following order, at
each site: (1) electrofishing, (2) push-kick netting, and
(3) seine netting. We employed these methods to target all major fish microhabitats so as to obtain comprehensive and unbiased fish communities (Supporting
Information).
Captured fishes were euthanized with MS-222, fixed in
10% formalin, and transferred to 75% ethanol for storage.
We used an analytical balance to measure the preserved
wet weight of each specimen and calculated the total fish
biomass in each sampling site. We identified all fish to
species (Kottelat 2013) and deposited specimens in Museum Zoologicum Bogoriense, Indonesia, and Lee Kong
Chian Natural History Museum, National University of
Singapore.
Instream Environmental Variables
In each site, we recorded 13 instream environmental
variables that could explain fish community responses:
littoral leaf litter cover, littoral root cover, volume of large
woody debris, substrate size, overstorey canopy cover,
stream width, stream depth, stream velocity, dissolved
oxygen concentration, water conductivity, dissolved nitrate concentration, dissolved iron concentration, and
water pH (Supporting Information).
Site- and Meso-Scale Forest Cover
We quantified forest cover in riparian zones of multiple
widths at 2 scales (site and meso-scale) by visually interpreting a true-color WorldView 2 satellite image at 0.5 m
resolution. Site scale was defined as the area immediately
adjacent to each 100 m sampling site. Meso-scale was
defined as the area adjacent to the 1000 m stream section
extending upstream from the downstream boundary of
each sampling site. We used ArcGIS 10.2 (ESRI, Redlands,
California) to measure site- and meso-scale forest cover at
different widths (10, 25, 50, 100, and 200 m) extending
outward from both sides of the stream.
Preconversion Species Richness in Continuous Forests
We were unable to sample continuous forest streams
because our study landscape (and much of lowland Central Kalimantan) is deforested. We therefore identified 5
streams that were described as “forest” or “forested” (but
otherwise physically similar to our RR and OP sites in
Conservation Biology
Volume 29, No. 5, 2015
Conserving Fishes in Oil-Palm Monocultures
1360
Figure 1. (a) Study site and fish sampling locations in Central Kalimantan Province, Indonesia, on the island of
Borneo (inset: black dot, location of study sites; solid gray shading, Mentaya drainage; gray shading with lines,
Kapuas drainage, for which the baseline fish community was reconstructed [see Methods]). Examples of stream
habitat types: (b) community forest riparian reserve (RRCF ) (Sungai Karuk); (c) logged forest riparian reserve
(RRLF ) (Sungai Samai); and (d) oil-palm (OP) (Sungai Balai).
terms of stream width, depth, substrate, and water color
or pH) in Tyson R. Roberts’ 1976 survey of the Kapuas
drainage, West Kalimantan (Roberts 1989), as preconversion controls (Supporting Information). These forest or
forested streams were likely embedded in the typical preconversion forest landscape—a continuous forest matrix
of largely logged or secondary forest, community forest
gardens, and primary forest patches—because the Kapuas survey predated large-scale oil-palm cultivation in
the province (Potter 2008).
Sampling lengths of the Kapuas sites were estimated
to be at least 100 m (and more likely 200 m) by
the researcher who performed the field work (T. R.
Roberts, personal communication). To ensure that the
Kapuas sites were comparable to our sites, we scaled
species richness values to the length of our sampling sites
Conservation Biology
Volume 29, No. 5, 2015
(100 m) by applying a richness versus sampling length
model developed for small lowland forest streams (dos
Anjos & Zuanon 2007) (Supporting Information). To assess the sensitivity of our results to the uncertainty in
sampling length, we conducted the analyses for a wide
range of values (100, 110, 125, 150, 200, and 300 m). Because the Kapuas fish communities were sampled using
ichthyocide, these sites were likely to be as (or more)
comprehensively sampled as our RR and OP sites and
were therefore adequate (or conservative) controls.
Functional Diversity
Functional diversity can be defined as the variation in
species traits that influence ecosystem function (Petchey
& Gaston 2006). We focused on one particular aspect
Giam et al.
of ecosystem function: energy flow (Violle et al. 2007).
For each species, we compiled 9 species traits hypothesized to affect energy flow in stream ecosystems through
their influence on a species’ life history, feeding strategy,
habitat use, and locomotion. The traits compiled from
the taxonomic literature (e.g., Roberts 1989), FishBase
(www.fishbase.org), and field observations were body
size, body shape, trophic position, mouth position, presence of jaw teeth, gregariousness, presence of barbels,
vertical position in water column, and air-breathing capability (Supporting Information).
We quantified functional diversity with 4 complementary metrics: functional group richness (FGR) (Petchey
& Gaston 2006), functional dispersion (FDis) (Laliberté
& Legendre 2010), functional evenness (FEve) (Villéger
et al. 2008), and Rao’s quadratic entropy (Rao’s Q) (BottaDukát 2005). FGR is calculated by delimiting ecologically
relevant functional groups following hierarchical clustering of species based on functional traits (Supporting Information). FDis is the average distance of individual species
to the community centroid in trait space (as projected by
principal coordinates analysis). FEve quantifies community evenness in trait space. Rao’s Q is similar to FDis
in that it describes community dispersion in trait space
by measuring the mean pairwise functional distances between species. We used the FD package (Laliberté &
Shipley 2011) in R 3.1.0 (R Core Team 2014) to calculate
these metrics.
Data Analyses
We fitted and analyzed generalized linear models (GLMs)
in a multimodel inferential framework (Burnham &
Anderson 2002) to examine whether local species richness differed among continuous forest, RRCF , RRLF , and
OP sites (Supporting Information).
We then investigated the mechanisms controlling local
species richness. We first identified the best instream
environmental correlates of local species richness by
comparing univariate GLMs (Supporting Information).
Following that we assessed possible mechanisms linking
presence of RR to local species richness via instream environmental variables (Supporting Information) by comparing relative support across candidate structural equation models (SEMs). For the best-supported model, we
assessed model adequacy with Shipley’s (2009) d-sep test
and calculated standardized path coefficients to quantify
relative signal strengths across paths (Grace et al. 2012).
In addition, we explored the spatial scale at which forest
cover best predicts local species richness (Supporting
Information).
We examined whether biomass differed among RRCF ,
RRLF , and OP sites by comparing candidate GLMs (Supporting Information). To determine whether RRCF , RRLF ,
and OP sites supported distinct species communities,
1361
we used the vegan package (Oksanen et al. 2014)
in R to perform nonmetric multidimensional scaling
(NMDS) and permutational multivariate analysis of variance (perMANOVA) (Anderson 2001). We performed
sample-based rarefaction in EstimateS (Colwell 2013) to
compare the total species richness supported by each
stream type.
Finally, we examined whether functional diversity
(FGR, FDis, FEve, and Rao’s Q) differed among continuous forest, RR, and OP sites by comparing candidate GLMs
(Supporting Information). In all GLM and SEM analyses,
we included models with stream width as a covariate
to control for its potential effect on species richness,
biomass, and functional diversity.
The sample size in our study (12–17 sites) was low
relative to terrestrial studies (e.g., Peh et al. 2006;
Koh 2008) but typical of riparian studies (Martin-Smith
1998; Jones III et al. 1999; Mercer et al. 2014) because sample sizes in the latter are generally limited by
the number of independent yet comparable streams in
a study landscape. To make accurate inferences from
a limited sample size, we fitted Bayesian GLMs and
SEMs (Grace et al. 2012), from which inferences on
model coefficients and effect sizes are exact and free
from large-sample assumptions (Kéry 2010) (Supporting
Information).
In each GLM and SEM analysis, we identified the bestsupported model with modified deviance information
criterion (DIC∗ ) which incorporates a larger penalty for
model complexity (Spiegelhalter et al. 2014) because
conventional DIC tends to overfit models when sample:parameter size ratio is low (Plummer 2008; Spiegelhalter et al. 2014). To assess goodness of fit of each GLM
and SEM submodel, we calculated R2 between observed
and predicted values (Grace et al. 2012). To assess model
adequacy, we checked observed values against posterior
means and 95% prediction intervals.
Data and example R code are in Supporting
Information.
Results
Local Species Richness
The best-supported GLM indicated that values for local
species richness in RRLF , RRCF , and continuous forest
sites were similar to each another and higher than values
for OP sites after controlling for stream width, assuming
the sampling length of the Kapuas continuous forest sites
was 200 m (R2 = 0.50) (Fig. 2 & Supporting Information).
Changing the sampling length assumption for the Kapuas
sites from 110–300 m did not change our results; the bestsupported model remained the same (R2 = 0.47–0.52)
(Supporting Information).
Conservation Biology
Volume 29, No. 5, 2015
Conserving Fishes in Oil-Palm Monocultures
1362
site to meso-scale) and riparian zone width (from 10 to
200 m) increased (Supporting Information).
Biomass, Community Composition, and Total Species
Richness
Figure 2. (a) Relationship between mean fish species
richness and stream width in continuous forest
(forest) and riparian reserve (RR) sites (solid line)
versus oil-palm plantation sites lacking riparian
reserves (OP) (dashed line) as indicated by the
best-supported model. (b) Model-fitted mean fish
biomass in forest and RR sites versus OP sites
(diamonds). Shaded areas are 95% prediction
intervals.
The best-supported GLM indicated that fish biomass was
similar in RRLF and RRCF sites and higher in RRLF and RRCF
sites than in OP sites (R2 = 0.39) (Fig. 2 & Supporting
Information).
Community composition was similar in RRLF and RRCF
sites (perMANOVA R2 = 0.13, p = 0.151). However, RR
sites (consisting of RRLF and RRCF sites pooled together)
were distinct from OP sites (perMANOVA R2 = 0.24,
p = 0.009). The NMDS ordination supported this result
(Fig. 4).
Total species richness (rarified to n = 3 sites) was
higher in RRLF (mean [SD] = 42 [2.4]) and RRCF sites
(40.8 [2.6]) than in OP sites (26.5 [2.7]) (Fig. 5).
Functional Diversity
For all functional diversity metrics, the best-supported
model indicated that functional diversity was similar in
RR sites and continuous forest sites and higher in RR sites
and continuous forest sites than in OP sites (Table 1 &
Supporting Information).
Mechanisms Maintaining Local Species Richness
Among instream environmental variables, species richness was positively correlated with substrate size, littoral
leaf litter cover, canopy cover, and stream depth and negatively correlated with dissolved nitrate concentration
and water conductivity (Supporting Information). Leaf
litter cover and substrate size were the 2 best predictors
of richness both before and after controlling for stream
width.
The best-supported SEM indicated that riparian reserves increased local species richness by increasing
leaf litter cover and maintaining substrate size; leaf litter cover exerted a stronger effect on species richness than substrate size (standardized path coefficients
= 0.55 and 0.30, respectively) when stream width
was controlled (Fig. 3 & Supporting Information). After accounting for indirect effects of substrate size
and leaf litter cover, riparian reserves had no direct
effect on species richness. The d-sep test confirmed
that the best-supported SEM was adequate (Supporting
Information).
Forest Cover, Local Species Richness, and Spatial Scales
Percent forest cover in the 10 m riparian zone directly
adjacent to the sampling site (site scale) best predicted
species richness. Predictive power of forest cover (R2
and DIC∗ ) showed a general decline as spatial scale (from
Conservation Biology
Volume 29, No. 5, 2015
Discussion
Our study is the first to indicate that retention of forest
patches along streams is effective in conserving aquatic
biodiversity in oil-palm plantations. At the local scale,
streams within forested riparian reserves in oil-palm plantations supported as many fish species as continuous
forests, but streams lacking forested riparian reserves
supported far fewer species. For example, our results
showed that a 5 m wide OP stream is expected to support, on average, 42% fewer species than a stream of the
same size in riparian reserves or continuous forests. At
the landscape scale, total species richness was 36% lower
in streams lacking riparian reserves than in streams with
riparian reserves.
How then does retention of riparian reserves in oilpalm plantations maintain fish species richness? From
our best-supported SEM, we inferred that littoral leaf litter cover and substrate size increased with the presence
of riparian reserves and that these 2 instream environmental variables in turn increased fish species richness.
The presence of riparian forest likely maintained quantity
and quality of leaf litter inputs at preconversion levels
(Pringle et al. 1988). Leaf litter supports fish communities by enhancing and spatially concentrating their food
resources—invertebrates, biofilm, and algae (Pringle
et al. 1988; Wallace et al. 1997). Leaf litter also provides
fishes with a cool, dark microhabitat and shelters them
Giam et al.
1363
Figure 3. Best-supported structural equation model of how riparian reserves drive local fish species richness via
instream environmental variables. (a) Causal diagram showing standardized path coefficients. Effect of riparian
reserves on (b) littoral leaf litter cover and (c) substrate size. Effects of (d) littoral leaf litter cover and (e) substrate
size on local species richness (diamonds, model-fitted means; shaded bars, 95% prediction intervals; black lines,
relationship between species richness versus [d] littoral leaf litter cover and [c] substrate size when all other
variables are constant at their mean values; data points for substrate size are jittered for clarity). Abbreviations:
RRLF , logged forest riparian reserve sites; RRCF , community forest riparian reserve sites; OP, oil-palm plantation
sites lacking riparian reserves.
from predators (Sazima et al. 2006). Fine niche partitioning in terms of space and food among species present in
leaf litter may also contribute to high species richness in
RR streams (Henderson & Walker 1990).
Besides providing leaf litter inputs, riparian reserves
likely reduce soil erosion from adjacent uplands and
therefore minimize fine substrate deposition into streams
(Jones III et al. 1999). Increased siltation and sedimentation reduce fish spawning grounds and zoobenthic food resources (Berkman & Rabeni 1987; Jones III
et al. 1999), thereby reducing fish species richness. Our
field observations corroborated this: species occurring
in streams with gravel (or larger) substrates (mostly benthic or benthopelagic species, for example, Glyptothorax
spp., Nemacheilus spp., Acantopsis dialuzona) did not
usually occur in clayey or silty streams.
Given the importance of riparian reserves for fish, the
scale at which forest cover influences species richness
needs to be determined. Forest cover in the 10 m zone
directly adjacent to the sampling sites best predicted richness patterns, most probably because this is the scale at
which terrestrial leaf litter enters the stream. However,
species richness remained positively correlated with
forest cover up to the maximum spatial scale investigated
(200 m riparian zone extending to 1 km upstream of
sampling site). Thus, while maintaining forest cover in
the first 10 m of the riparian zone is arguably the most
important step in conserving fish in oil-palm plantations,
retaining forest cover over larger spatial scales appears to
also increase species richness, likely by decreasing sedimentation and chemical runoff and maintaining stream
temperatures.
The community composition of RR streams was distinct from that of OP streams, which is consistent with
our finding that removing riparian reserves decreased
species richness through reduction of leaf litter and
coarse substrate microhabitats. Species that specialize in
foraging or hiding in these microhabitats (e.g., loaches,
small catfishes) were predictably extirpated in the absence of RR. Pelagic species that do not appear to specialize in particular benthic microhabitats occurred in both
OP and RR streams (e.g., Puntigrus navjotsodhii and
Rasbora elegans) or only in OP streams (Trigonopoma
pauciperforatum).
Streams protected by riparian reserves maintained
functional diversity comparable to the level observed in
Conservation Biology
Volume 29, No. 5, 2015
Conserving Fishes in Oil-Palm Monocultures
1364
Figure 4. Fish community composition in logged
forest riparian reserve sites (RRLF ), community forest
riparian reserve sites (RRCF ), and oil-palm plantation
sites lacking riparian reserves (OP) as described by
nonmetric dimensional scaling ordination of sites. See
Supporting Information for corresponding ordination
of fish species.
continuous forest, whereas OP streams lacking RR exhibited a reduction in functional diversity. FGR was higher in
RR than in OP streams. We obtained the same result with
richness-independent metrics, FDis and Rao’s Q, which
suggests higher functional diversity in RR streams did not
arise just from having more species. Functional groups
rare or absent in OP streams included surface-dwelling
terrestrial insectivores (e.g., Hemirhamphodon phaiosoma) and species associated with specific benthic microhabitats such as gregarious benthic invertivores (e.g.,
Pangio spp. associated with leaf litter; Acanthoposoides
spp. associated with coarse substrate). The removal of
riparian reserves also reduced FEve as pelagic, schooling cyprinids dominated the species assemblage alongside a few functionally different species (e.g., barbelled
[Hemibagrus capitulum] and air-breathing [Channa
Figure 5. Total fish species richness in logged forest
riparian reserve sites (RRLF ), community forest
riparian reserve sites (RRCF ), and oil-palm plantation
sites lacking riparian reserves (OP). Shaded areas are
95% confidence intervals.
striata] predators). Together, these results suggest that
functional diversity losses in OP streams were driven
by the loss of benthic microhabitats and terrestrial food
resources. If these functional groups do not recover as
riparian forest regenerates, terrestrial energy inputs (via
benthic food chains and consumption of terrestrial insects by fish) may be attenuated, potentially decreasing
secondary fish production.
Fish biomass was similar in logged forest and community forest RR streams and higher at these streams
than in OP streams. This likely resulted from a combination of greater food (energy) availability and a
greater diversity of species with complementary niches,
conditions that maximized energy transfer and increased fish production in riparian reserves sites relative to OP sites. By retaining riparian reserves land
managers can ensure that fish for subsistence use,
an important ecosystem service, is maintained at high
levels.
Across the multiple biodiversity metrics we investigated, riparian reserves comprising in community forests
Table 1. Relationship between functional diversity metrics of fish communities and stream types (i.e., continuous forest streams, oil-palm plantation
streams located within forested riparian reserves, and oil-palm plantation streams lacking forested riparian reserves) based on the best-supported
generalized linear model.
Stream type, mean (SD)
Functional diversity metricsa
FGR
FDis
Rao’s Q
FEve
a Abbreviations:
b Abbreviations:
Continuous forest
Forested riparian reserve
Oil-palm
Inferenceb
R2
9.20 (1.92)
0.32 (0.01)
0.11 (0.01)
0.95 (0.02)
11.38 (3.62)
0.33 (0.02)
0.11 (0.01)
0.95 (0.02)
6.75 (3.30)
0.27 (0.08)
0.09 (0.04)
0.90 (0.04)
(forest = RR) > OP
(forest = RR) > OP
(forest = RR) > OP
(forest = RR) > OP
0.63
0.28
0.23
0.48
FGR, functional group richness; FDis, functional dispersion; FEve, functional evenness; Rao’s Q, Rao’s quadratic entropy.
RR, forested riparian reserve sites; OP, oil-palm plantation sites lacking riparian reserves.
Conservation Biology
Volume 29, No. 5, 2015
Giam et al.
with active human use performed as well as riparian
reserves comprising currently unused but previously
logged forests. This suggests limited amounts of extractive use do not compromise the utility of riparian reserves for protecting fish, which lowers the opportunity
cost of creating such reserves. However, we do not yet
know if the efficacy of riparian reserves will persist over
time. Given that small tropical freshwater fishes have a
typical lifespan of 2–5 years (Ng & Tan 1997) and that
we captured juveniles as well as adults of many species
across all sites (X.G., unpublished data), fish communities were likely stabilized when we conducted our surveys (6 years after plantation establishment). Nevertheless, our results provide a much-needed baseline for
future studies on long-term efficacy of maintaining or
restoring riparian reserves.
Our study would have benefitted from a larger number
of sampling sites. Unfortunately, this was not possible because we sampled the maximum number of comparable
and independent sites in our study landscape. However,
to ensure that the limited replication did not invalidate
our analyses, we fitted Bayesian models that are free from
large-sample assumptions. Moreover, the effect sizes in
all analyses were medium to large (R2 of best models
= 0.23–0.87), giving us confidence in our findings. A
potential pitfall is that the model selection metric, DIC,
becomes less accurate and tends to overfit models when
sample sizes are low. We alleviated this problem by using
a modified metric (DIC∗ ) that assigns a larger penalty
for model complexity (Spiegelhalter et al. 2014). Overall,
we are confident that our model inferences are robust;
however, further research, ideally incorporating a larger
number of sampling sites over multiple drainages and
regions (e.g., Borneo, Sumatra, and Peninsular Malaysia),
would help confirm our findings as well as examine alternative mechanisms by which RR can conserve fish
communities (e.g., by reducing chemical runoff from
pesticides).
Although the retention of riparian reserves is mandated
by law in Indonesia and Malaysia, noncompliance is common owing to poor enforcement (Lajiun 2013) and, possibly, legal ambiguity. For example, in Indonesia, small and
large rivers are mandated to have 50 and 100 m buffers
of natural vegetation, respectively (Republic of Indonesia
1990). Yet, definitions of small and large rivers and natural vegetation are lacking; thus, oil-palm growers have
wiggle room. Relevant governmental agencies should
clarify legislative language pertaining to river width, vegetation type, and type of production landscapes in which
riparian reserves are required.
Beyond better definition and enforcement of laws
by governments, international and national multistakeholder regulatory groups (e.g., RSPO and Indonesia Sustainable Palm Oil) should ensure that member companies
maintain riparian reserves. Downstream participants in
the palm-oil supply chain such as purchasers and end
1365
users could drive changes in environmental performance
of upstream growers. By purchasing only palm oil originating from plantations that retain riparian reserves,
downstream participants can create market-based incentives to preserving riparian forests.
Our study suggests that a land-sharing strategy that
involves retaining forests along streams is effective in
conserving freshwater fish biodiversity in oil-palm plantations. Streams with forested riparian reserves do not
just hold more fish species than those that lack riparian reserves, they also support higher fish biomass and
functional diversity, which are important for sustaining
human livelihoods and ecosystem function, respectively.
Our results are globally relevant because oil-palm monoculture is expanding in both the Neotropics (Lees &
Vieira 2013) and Afrotropics (Wich et al. 2014). We therefore urge governments, regulatory groups, growers, and
downstream participants in the oil-palm supply chain to
make conservation of forested riparian reserves a key
policy objective.
Acknowledgments
This research was principally funded by a Wildlife Reserves Singapore grant and the Mary and Randall Hack
’69 Award to X.G., and a High Meadows Foundation grant
to D.S.W. Rufford Foundation and Idea Wild provided
additional funding. The National Museum of Natural History, Smithsonian Institution, supported the publication
of this paper. We are grateful to T.R. Roberts for providing
information on the Kapuas sites. We thank L.T. Gan, T.S.
Chock, management, and staff of Musim Mas Group for
research permission and field support, RISTEK for granting a research permit (No: 20/EXT/SIP/FRP/SM/III/2013),
and LIPI for coordinating research. We benefitted from
discussions with J. Green and A. Yee. We also thank the
handling editor, M. Jones, and 4 anonymous reviewers
whose comments greatly improved this manuscript. This
research is approved by Princeton University IACUC (Protocol 1861).
Supporting Information
Dendrogram of functional groups (Appendix S1), relationship between instream environmental variables and
species richness (Appendix S2), relationship between
forest cover and species richness (Appendix S3), species
ordination (Appendix S4), summary of instream environmental variables (Appendix S5), description of functional
traits (Appendix S6), functional trait values (Appendix
S7), functional groups (Appendix S8), candidate model
sets (Appendices S9–S14), model rankings (Appendices
S15–S17, S19–21), d-sep test (Appendix S18), physical
characteristics and species richness of continuous forest
Conservation Biology
Volume 29, No. 5, 2015
1366
sites (Appendix S22), fish sampling methods (Appendix
S23), instream environmental variables (Appendix S24),
continuous forest sites: description and richness scaling
(Appendix S25), Bayesian model settings (Appendix S26),
and data and code (Appendix S27) are available online.
The authors are solely responsible for the content and
functionality of these materials. Queries (other than absence of the material) should be directed to the corresponding author.
Literature Cited
Allan JD, Flecker AS. 1993. Biodiversity conservation in running waters.
BioScience 43:32–43.
Anderson MJ. 2001. A new method for non-parametric multivariate
analysis of variance. Austral Ecology 26:32–46.
Balmford A, Green R, Phalan B. 2012. What conservationists need to
know about farming. Proceedings of the Royal Society B 279:2714–
2724.
Berkman HE, Rabeni CF. 1987. Effect of siltation on stream fish communities. Environmental Biology of Fishes 18:285–294.
Botta-Dukát Z. 2005. Rao’s quadratic entropy as a measure of functional
diversity based on multiple traits. Journal of Vegetation Science
16:533–540.
Burnham KP, Anderson DR. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Springer, New
York.
Casatti L, Teresa TB, Gonçalves-Souza T, Bessa E, Manzotti AR,
Gonçalves CdS, Zeni JdO. 2012. From forests to cattail: How does
the riparian zone influence stream fish? Neotropical Ichthyology
10:205–214.
Colwell RK. 2013. EstimateS: Statistical estimation of species richness
and shared species from samples. Version 9.1.
Core R Team. 2014. R: A Language and Environment for Statistical
Computing. Vienna, Austria, R Foundation for Statistical Computing.
dos Anjos MD, Zuanon J. 2007. Sampling effort and fish species richness in small terra firme forest streams of central Amazonia, Brazil.
Neotropical Ichthyology 5:45–52.
Dudgeon D. 2000. The ecology of tropical Asian rivers and streams in
relation to biodiversity conservation. Annual Review of Ecology and
Systematics 31:239–263.
Edwards DP, Hodgson JA, Hamer KC, Mitchell SL, Ahmad AH, Cornell SJ, Wilcove DS. 2010. Wildlife-friendly oil palm plantations
fail to protect biodiversity effectively. Conservation Letters 3:236–
242.
Fayle TM, Turner EC, Snaddon JL, Chey VK, Chung AYC, Eggleton
P, Foster WA. 2010. Oil palm expansion into rain forest greatly
reduces ant biodiversity in canopy, epiphytes and leaf-litter. Basic
and Applied Ecology 11:337–345.
Gibson L, et al. 2011. Primary forests are irreplaceable for sustaining
tropical biodiversity. Nature 478:378–381.
Government of Malaysia. 1965. National Land Code (No. 56 of 1965).
Grace JB, Schoolmaster Jr DR, Guntenspergen GR, Little AM, Mitchell
BR, Miller KM, Schweiger EW. 2012. Guidelines for a graph-theoretic
implementation of structural equation modeling. Ecosphere 3,
art73. doi: http://dx.doi.org/10.1890/ES12-00048.1.
Gray CL, Lewis OT, Chung AYC, Fayle TM. 2015. Riparian reserves
within oil palm plantations conserve logged forest leaf litter ant
communities and maintain associated scavenging rates. Journal of
Applied Ecology 52:31–40.
Gray CL, Slade EM, Mann DJ, Lewis OT. 2014. Do riparian reserves
support dung beetle biodiversity and ecosystem services in oil palmdominated tropical landscapes? Ecology and Evolution 4:1049–
1060.
Conservation Biology
Volume 29, No. 5, 2015
Conserving Fishes in Oil-Palm Monocultures
Henderson PA, Walker I. 1990. Spatial organization and population
density of the fish community of the litter banks within a central
Amazonian blackwater stream. Journal of Fish Biology 37:401–411.
Jones III EBD, Helfman GS, Harper JO, Bolstad PV. 1999. Effects of
riparian forest removal on fish assemblages in Southern Appalachian
streams. Conservation Biology 13:1454–1465.
Kéry M. 2010. Introduction to WinBUGS for ecologists. Academic Press,
Burlington.
Koh LP. 2008. Can oil palm plantations be made more hospitable for
forest butterflies and birds? Journal of Applied Ecology 45:1002–
1009.
Kottelat M. 2013. The fishes of the inland waters of Southeast Asia: a
catalogue and core bibliography of the fishes known to occur in
freshwaters, mangroves and estuaries. Raffles Bulletin of Zoology
Supplement 27:1–663.
Lajiun J. 2013. Enforce law on oil palm plantation encroachment. The Borneo Post Online. Available from http://www.
theborneopost.com/2013/11/12/enforce-law-on-oil-palmplantation-encroachment-masidi/ (accessed April 2014).
Laliberté E, Legendre P. 2010. A distance-based framework for measuring functional diversity from multiple traits. Ecology 91:299–305.
Laliberté E, Shipley B. 2011. FD: measuring functional diversity from
multiple traits, and other tools for functional ecology. R package
v.1.0–11.
Lees AC, Vieira IC. 2013. Forests: oil palm concerns in Brazilian Amazon.
Nature 497:188.
Lorion CM, Kennedy BP. 2009. Riparian forest buffers mitigate the
effects of deforestation on fish assemblages in tropical headwater
streams. Ecological Applications 19:468–479.
MacKinnon K, Hatta G, Halim H, Mangalik A. 1997. The ecology of
Kalimantan. Periplus Editions, Hong Kong.
Martin-Smith KM. 1998. Relationships between fishes and habitat in
rainforest streams in Sabah, Malaysia. Journal of Fish Biology 52:458–
482.
Mercer EV, Mercer TG, Sayok AK. 2014. Effects of forest conversion
to oil palm on freshwater macroinvertebrates: a case study from
Sarawak, Malaysia. Journal of Land Use Science 9:260–277.
Millennium Ecosystem Assessment. 2005. Current state and trends assessment. Island Press, Washington D.C.
Naiman RJ, Décamps H. 1997. The ecology of interfaces: riparian zones.
Annual Review of Ecology and Systematics 28:621–658.
Ng PKL, Tan HH. 1997. Freshwater fishes of Southeast Asia: potential for
the aquarium fish trade and conservation issues. Aquarium Sciences
and Conservation 1:79–90.
Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O’Hara RB,
Simpson GL, Solymos P, Stevens MHH, Wagner H. 2014. Vegan:
community ecology package. R package v. 2.0–10.
Peh KS-H, Sodhi NS, De Jong J, Sekercioglu CH, Yap CA-M, Lim SL-H.
2006. Conservation value of degraded habitats for forest birds in
southern Peninsular Malaysia. Diversity and Distributions 12:572–
581.
Petchey OL, Gaston KJ. 2006. Functional diversity: back to basics and
looking forward. Ecology Letters 9:741–758.
Phalan B, Onial M, Balmford A, Green RE. 2011. Reconciling food production and biodiversity conservation: land sharing and land sparing
compared. Science 333:1289–1291.
Plummer M. 2008. Penalized loss functions for Bayesian model comparison. Biostatistics 9:523–539.
Potter L. 2008. Dayak resistance to oil palm plantations in West
Kalimantan, Indonesia. 17th Biennial Conference of the Asian
Studies Association of Australia. Available from http://artsonline.
monash.edu.au/mai/files/2012/07/lesleypotter.pdf (accessed April
2014).
Pringle CM, Naiman RJ, Bretschko G, Karr JR, Oswood MW, Webster
JR, Welcomme RL, Winterbourn MJ. 1988. Patch dynamics in lotic
systems: the stream as a mosaic. Journal of the North Americal
Benthological Society 7:503–524.
Giam et al.
Putz PE, Romero C. 2014. Futures of tropical forests (sensu lato).
Biotropica 46:495–505.
R Core Team. 2014. R: A Language and Environment for Statistical
Computing. Vienna, Austria, R Foundation for Statistical Computing.
Republic of Indonesia. 1990. Keputusan Presiden No. 32 Tahun 1990.
Roberts TR. 1989. The freshwater fishes of Western Borneo (Kalimantan
Barat, Indonesia). Memoirs of the California Academy of Sciences
14:1–210.
Rowe RK, Smith J, Quinn J, Boothroyd I. 2002. Effects of logging with
and without riparian strips on fish species abundance, mean size,
and the structure of native fish assemblages in Coromandel, New
Zealand, streams. New Zealand Journal of Marine and Freshwater
Research 36:67–79.
Sazima I, Carvalho LN, Mendonça FP, Zuanon J. 2006. Fallen leaves on
the water-bed: diurnal camouflage of three night active fish species
in an Amazonian streamlet. Neotropical Ichthyology 4:119–122.
Shipley B. 2009. Confirmatory path analysis in a generalized multilevel
context. Ecology 90:363–368.
Sodhi NS, Koh LP, Brook BW, Ng PKL. 2004. Southeast Asian biodiversity: an impending disaster. Trends in Ecology and Evolution
19:654–660.
Spiegelhalter DJ, Best NG, Carlin BP, Van Der Linde A. 2014. The
deviance information criterion: 12 years on. Journal of the Royal
Statistical Society B 76:485–493.
1367
Sweeney BW, Newbold JD. 2014. Streamside forest buffer width needed
to protect stream water quality, habitat, and organisms: a literature review. Journal of the American Water Resources Association
50:560–584.
Villéger S, Mason NW, Mouillot D. 2008. New multidimensional functional diversity indices for a multifaceted framework in functional
ecology. Ecology 89:2290–2301.
Violle C, Navas ML, Vile D, Kazakou E, Fortunel C, Hummel I, Garnier
E. 2007. Let the concept of trait be functional! Oikos 116:882–892.
Wallace JB, Eggert SL, Meyer JL, Webster JR. 1997. Multiple trophic
levels of a forest stream linked to terrestrial litter inputs. Science
277:102–104.
Wich SA, Garcia-Ulloa J, Kuhl HS, Humle T, Lee JSH, Koh LP. 2014. Will
oil palm’s homecoming spell doom for Africa’s great apes. Current
Biology 24:1–5.
Wilson EO. 1988. The current state of biological diversity. Pages 1-18
in Wilson EO, editor. Biodiversity. National Academy Press, Washington, D.C.
Wilcove, DS, Giam X, Edwards DP, Fisher B, Koh LP. 2013. Navjot’s
nightmare revisited: logging, agriculture, and biodiversity in Southeast Asia. Trends in Ecology & Evolution 28:531–540.
WWF and TNC (World Wildlife Fund and The Nature Conservancy).
2013. Freshwater ecoregions of the world (FEOW). Available from
www.feow.org (accessed January 2014).
Conservation Biology
Volume 29, No. 5, 2015