Content uploaded by Jia Huan Liew
Author content
All content in this area was uploaded by Jia Huan Liew on Jul 02, 2018
Content may be subject to copyright.
Ecol Freshw Fish. 2 018 ;1–12. wileyonlinelibrary.com/journal/eff
|
1
© 2018 John Wiley & Sons A /S.
Publish ed by John W iley & Sons Ltd
1 | INTRODUCTION
Fresh waters are some of the most threatened habitats globally with
approximately 10,000–20,000 species thought to be imperilled by
anthropogenic stressors (Strayer & Dudgeon, 2010). Some com-
monly documented stressors include eutrophication (Smith, Joye,
& Howarth, 2006; Vilmi, Karjalainen, Landeiro, & Heino, 2015),
overfishing (Allan et al., 2005), biological invasions (Correa, Bravo, &
Hendry, 2012; Gibbs, Shields, Lock, Talmadge, & Farrell, 2008) and
habitat change/modification (Didham, Tylianakis, Gemmell, Rand,
& Ewers, 2007; Giam et al., 2012). Of these, habitat modification
is particularly detrimental because it results in multiple, synergistic
Received:29December2017
|
Revised:3M ay2018
|
Accepted:4May2018
DOI :10 .1111/ef f.12419
ORIGINAL ARTICLE
Contrasting changes in freshwater fish assemblages and food
webs follow modification of tropical waterways
Jia Huan Liew1 | Xingli Giam2 | Esther Clews3 | Kenrick Y. W. Tan1 |
Heok Hui Tan4 | Zi Yi Kho4 | Darren C. J. Yeo1
1Depar tment of B iological Scie nces, National
University of Singapore, Singapore,
Singapore
2Department of Ecology and Evolutionary
Biology, The University of Tennessee,
Knoxville, Tennessee
3TropicalMarineScienceInstitute,N ational
University of Singapore, Singapore,
Singapore
4Lee Kong Chian Natur al History
Museum ,NationalUniver sityofSingapore,
Singapore, Singapore
Correspondence
Darren C . J. Yeo, Department of Biolog ical
Science s, National University of Singapore,
14ScienceDr ive4,Singapore117543 ,
Singapore.
Email: dbsyeod@nus.edu.sg
Funding information
National Research Foundation and the
Economic Development Board, Grant/Award
Number: COY-15-EWI-RCFSA/N197-1;
Public Utilitie s Board - Si ngapore, Grant/
AwardNumber:R-154-0 00 -619-490
Abstract
Fresh waters are increasingly threatened by flow modification. Knowledge about the
impacts of flow modification is incomplete, especially in the tropics where ecological
studies are only starting to emerge in recent years. Using presence/absence data
dated approximately four decades apart (~1966 to ~2010) from 10 tropical rivers, we
assessed the changes in freshwater fish assemblage and food web after flow modifi-
cation. The sites were surveyed with methods best suited to habitat conditions (e.g.,
tray/push netting for low- order forest streams, visual surveys for canalised rivers and
net casting for impounded rivers). With the presence/absence data, we derived and
compared six measures of fish assemblage and food web structure: species richness,
proportion of native species, overall functional diversity, native functional diversity,
food web complexity and maximum trophic level. We found that changes in commu-
nity assemblage and food web structure were not generalisable across modification
regimes. In canalised sites, species richness and maximum trophic levels were lower
in the second time period while the opposite was true for impounded sites. However,
proportion of native species was consistently lower in the second time period across
modification regimes. Changes in fish assemblages and food webs appear to be
driven by species turnover. We recorded 79 cases of site- specific extirpation and 117
cases of site- specific establishment. Our data further suggest that turnover in as-
semblage is again contingent on flow- modification regime. While the process was
stochastic in canalised rivers, benthopelagic species were more likely to be extir-
pated from impounded rivers where species lost were replaced by predominantly
alien fish taxa.
KEYWORDS
alien species, anthropogenic impacts, canalisation, functional diversity, impoundment, trophic
ecology
2
|
LIEW Et aL.
stressors (Liew, Tan, & Yeo, 2016; Liu, Wang, & C ao, 2012; Vitule,
Skora, & Abilhoa, 2012). The rate of habitat modification in fresh
waters has accelerated in recent years (Gopal & Ghosh, 2009;
MillenniumEcosystemAssessment2005), and thistrend isunlikely
to abate with increasing levels of urbanisation, considering the social
and economic importance of fresh waters.
Fresh waters are commonly managed via flow modification
(sensu Dudgeon, 2014), which involves the construction of river
impoundments or canals. Impoundments are known to cause ex-
tirpation, particularly among obligate riverine species (Liew, Tan
etal.,2016;Quinn&Kwak,2013;Zhang,Gao,Wang,&Cao,2015).
Conversely, canalisation impacts freshwater communities via loss
of refugia (Millidine, Malcolm, Gibbins, Fr yer, & Youngson, 2012;
Tockner, Pusch, Borchardt, & Lorang, 2010) and the alteration of
flow varia bility and/or seas onality (Dud geon, 2014). Althoug h the
effects of flow modification on freshwater communities have been
researched extensively (Melcher, Ouedraogo, & Schmutz, 2012;
Penczak, Agostinho, Gomes, & L atini, 2009; Quinn & Kwak, 2013;
Simoes et al., 2015), coverage of scientific knowledge is uneven.
Prominently, tropical fresh waters are understudied (Magurran &
Queiroz,2010)inspiteofacceleratingthreatsfromrapidlygrowing
humanpopulationsintheregion(Edelmanetal.,2014).
Gaps in the spatial coverage of flow- modification research are
important to address because nuances in spatial trends of ecological
processes can confound management/conservation efforts (Boulton
et al., 2008). For example, impacts from the removal of riparian veg-
etation (and associated allochthonous coarse particulate organic
matter) on shredder- scarce tropical communities are not likely to
be consistent with predictions made by classical ecological studies
conducted in temperate water bodies (Boyero, Ramirez, Dudgeon,
& Pearson, 2009; Goncalves, Graca, & Callisto, 20 06). Spatial con-
founders, however, are not solely a function of latitude (Boulton
et al., 2008). Historical flow regimes (e.g., flood pulses) have also
been shown to influence the resilience of ecological processors to
stressors(Lytle&Poff,2004;Rollsetal.,2013),thusfurtherconfus-
ing management/conservation decisions.
Anthropogenic stressors impact more than just species commu-
nities. Evidence suggesting that human activities adversely affect
ecosystem functions is accruing in contemporary limnological litera-
ture (VanCappellen & Maavara, 2016; Vaughn, 2010). As ecosystem
functions are closely linked with trophic interactions ( Thompson et al.,
2012), consequences of disruptions to the former can be reflected in
changesinfoodwebstructure(Duedoro,Box,Vazquez-Luis,&Arroyo,
2014).Unfor tunately,our knowledge about the impacts of flow mod-
ification on food webs is incomplete. Existing data suggest that some
structural changes do result from impoundments. Known changes
include increased food web dispersion (Mercado-Silva, Helmus, &
VanderZanden,2008)andfoodchainlength(Hoeinghaus,Winemiller,
& Agostino, 2008) although little else has been documented. On the
other hand, the consequences of canalisation on food web structure
remain largely unresolved.
In this study, we address these knowledge gaps by assessing
the changes in the species assemblage and food web structure in
modified tropical waterways. To this end, we analysed fish pres-
ence/absence data from two time periods (approximately four de-
cades apart) representing conditions before and after significant
anthropogenic flow modification. Our primary research questions
were as follows: (a) What are the changes in tropical fish assem-
blages associated with various flow- modification regimes?; (b) How
do the different flow- modification regimes affect food web struc-
ture?; and (c) What are the mechanisms underlying these changes?
We addressed questions (b) and (c) using food web and functional
diversity metrics calculated from literature- derived functional trait
and trophic ecolog y (diet) data.
2 | METHODS
2.1 | Study sites and design
We assessed data from 10 sites in the tropical island nation of
Singapore (Table 1; Figure 1). Fish assemblage data were collected
intwotimeperiods,1957–1964and2006–2010.Betweenthetime
periods, Singapore’s inland water bodies and coasts underwent
significant changes as a consequence of urbanisation and coastal
reclamation respectively (Figure 1). Among our study sites, four
were canalised (i.e., Punggol, Serangoon, Simpang and Kallang riv-
ers) and five were impounded (i.e., Jurong, Kranji, Lower Seletar,
Tengeh and Upper Seletar) since the first time period. One study
site(i.e., MacRitchie)wasanexception in that it was impounded in
the nineteenth century and remained relatively unchanged between
time periods, thus serving as a reference point (henceforth referred
to as lentic control). We grouped the sites according to their flow-
modification regimes (or treatments).
2.2 | Data collection
Fish assemblage (i.e., species presence/absence) data for the
first time period were collated from sur veys conducted by Alfred
(1966), in which the author described fish communities in all the
major rivers (and one reservoir) in Singapore. Species lists from
this publication comprise fish specimens collected by the author
betwee n 1957and 1964, a s well as records of f ish occurren ces
in Singapore waterways from various natural history museums
(i.e., National Museumof Singapore (zoological collections now
with the Lee Kong Chian Natural History Museum, National
University of Singapore), Zoologisch Museum Amsterdam, the
former B ritish Muse um of Natural Hi story (now N atural Hist ory
Museum, London) andthe former Rijksmuseum van Natuurlijke
HistorieLeiden(nowNationaalNatuurhistorischMuseum)).Alfred
(1966) adopted a classical natural history approach with a focus
on taxonomic descriptions (i.e., identification keys) and the pres-
ence/absence of fish species in various freshwater bodies, so little
information pertaining to sampling protocol was available. As was
typical of early taxonomic explorations, specimen collection was
likely to be ad hoc with a mix of sampling methods including rec-
tangular tray/push nets, seine nets, cast nets and fish traps (pers.
|
3
LIEW Et a L.
comm. with fish curators of the Lee Kong Chian Natural History
Museum).Alfred’sspecieslistsweresupplementedbyspecimens
contributed to natural histor y museums (listed above) by other
researchers. From these information sources, we recorded 40
unique species occurring at our study sites in the first time period
(Appendix A).
Data for the second time period were collected with two differ-
ent approaches. In impounded (n = 5) and lentic control (n = 1) sites,
fish assemblage data were obtained from a broad- based survey of
biodiversity in Singapore’s reservoirs (Ng & Tan, 2010). Each lentic
study site (i.e., lentic control and impounded rivers) was surveyed
over 12–18 months, during which three to six sampling occasions
were conducted, per site. All sample occ asions spanned an average
of 1 week. The total sur vey duration and frequency of sample occa-
sions were dictated by logistical constraints (e.g., boat availability)
stemming from the need to coordinate survey effor ts with reser-
voir managers and administrators. In each week- long sampling oc-
casion, a combination of fish collection methods (i.e., trapping, net
casting, trawling, deployment of baited fish hooks) were deployed
at two randomly selected littoral transects (<5 m from shore) of ap-
proximately 500 m. The depth of these transect s varied; thus, the
complementary use of multiple fish collection methods in each sam-
pling occasion was necessar y to cover the entire water column. For
example, fish traps were more effective for benthic/benthopelagic
taxa while net casting were better suited for the c apture of pelagic
species. Sampling effor t (i.e., t ype of fish collection method and sur-
vey intensity) was kept consistent across sampling occasions.
Fish assemblage data from canalised sites (n = 4)werecollected
through visual sur veys in 2010. These comprised surveys over 50–
400m stretches(dependingon accessibility) atalong three to five
randomly selected point s along each drainage. To maximise species
detection, both diurnal and nocturnal sur veys were conducted by
groups of three to four observers with the aid of binoculars, cameras
withzoomlensandtorcheswhennecessary.Wherespeciesidenti-
fication was uncer tain, photographs were acquired for confirmation
against reliable identification guides such as Baker and Lim (2012).
As thesehabitatssites were narrow,shallow (<0.3m) andspatially
homogenous (Appendix F), detection rates were likely to be high
(McNew & Handel, 2015). Moreover, water turbidity at t he sites
was generally low. We note, however, that some species may have
been missed in our visual surveys because a small subset of spe-
cies encountered in our study are cryptic—Rhinogobius giurinus and
TABLE1 Summary of study sites and their conditions in the first and second time periods respec tively. Wasteland refers to unmanaged
vegetationcomprisingamixofoldandyoungsecondaryforests,aswellasscrubland.Sizeofstudysitesinthesecondtimeperiodwas
estimated from data made available under the Singapore Open Data License (2018)
Treatment type Site(s) Year modified Fir st time period (~1966) Second time period (~2010)
Lentic control MacRitchie 1868 Impounded, landlocked, within nature
reserve.
Impounded, landlocked, within nature
reser ve. Surface area = 0.9 km2, shore
length=13.8km.
Canalised site Kallang Unknown Unmodified, marine connection, mixed
urban and wasteland matrix.
Canalised, no direct marine connection,
urban mat rix. Approximate length = 8.9 km.
Punggol Unknown Unmodified, marine connection, mixed
agriculture and wasteland matrix.
Canalised, marine connection, mixed urban
and wasteland matrix. Approximate
length = 5.7 km.
Serangoon Unknown Unmodified, marine connection, mixed
rural village and wasteland matrix.
Canalised, marine connection, mixed urban
and wasteland matrix. Approximate
length = 6.7 km.
Simpang Unknown Unmodified, marine connection, mixed
rural village and wasteland matrix.
Canalised, no direct marine connection,
mixed urban and wasteland matrix.
Approximate leng th = 2.9 km.
Impounded site Jurong 1971 Unmodified, marine connection, mixed
rural village and wasteland matrix.
Impounded, landlocked, within recreational
park. Surface area = 1.7 km2, shore
length = 6.2 km.
Kranji 1971 Unmodified, marine connection, mixed
rural village and wasteland matrix.
Impounded, landlocked, mixed urban and
wasteland matrix. Surface area = 5.0 km2,
shorelength=53.7km.
Lower Seletar 1983 Unmodified, marine connection, mixed
rural village and wasteland matrix.
Impounded, landlocked, within nature
reser ve.Surfacearea=3.2km2, shore
length = 17.5 km.
Tengeh 1977 Unmodified, marine connection, mixed
rural roads and wasteland matrix.
Impounded, landlocked, within protected
forest .Surf acearea=1.3km2, shore
length = 9.5 km.
Upper Seletar 1967 Unmodified, landlocked, mixed urban
and wasteland matrix.
Impounded, landlocked, within nature
reser ve.Surfacearea=3.1km2, shore
length=33.7km.
4
|
LIEW Et aL.
Nemacheilus selangoricus being prominent examples (the former was
recorded from Kallang before canalisation).
One issue we faced in our approach to data collection was incon-
sistencies in survey methodology, which was unavoidable consider-
ing that methods were chosen to best suit habitat conditions (e.g.,
waterbody depth,habitattypeand heterogeneity).Moreover,the
independent sampling for Alfred’s (1966) historically significant data
set and the modern/current data set (Ng & Tan, 2010; present study)
wereconducted almost40years apart; thus,protocol standardisa-
tion was not feasible. To minimise comparative bias, we analysed
only (qualitative) species presence/absence data. We also assessed
the sampling completeness of all three data sources (i.e., species lists
from Alfred (1966), fish assemblage data from a broad- based survey
of Singapore’s reservoirs (Ng & Tan, 2010), and species lists from the
present visual surveys of canalised sites) with an incidence- based
coverage estimator (Chao & Jost, 2012) using the iNEXT*2.0.12
statistical package(Hsieh, Ma,& Chao,2016).We found thatsam-
pling completeness was relatively high, ranging between 81.7% and
94.5%(AppendixG).
We identified a total of 82 unique fish species from all the water
bodies in both time periods (Appendix A). For all species present,
we recorded functional traits from FishBase (Froese & Pauly, 2011).
These comprised maximumsize, bodyshape, trophic breadth,tro-
phic guild, trophic level, water column preference, resilience and re-
productive strategy (e.g., mouth brooder, nest builder) (Appendix B).
We examined changes in functional composition across all treatment
types qualitatively by constructing a Gower’s dissimilarity cluster
dendrogram (Gower, 1971). Finally, we summarised overall func-
tional diversity per site per time period using the Rao’s quadratic
entropy (Q) index which quantifies mean pairwise functional dis-
tance between all possible species permutations, and is therefore
indicative of functional trait dispersion (Botta- Dukat, 20 05). We also
calculated a modified version of Q in which all non- native species
were excluded. This modified index was termed native functional
diversity (Qn).
Hypothe sised food webs associated with e ach study site per time
period we re constructed using protocols descr ibed in Liew, Carrasco,
Tan, and Yeo (2016). Briefly, this involved the elucidation of trophic
linksusingdietandsizedatafromliterature.Thesewebsweresum-
marised with indices reflecting complexity (i.e., connectance) and
maximum trophic level. The former is a measure of the proportion of
realised links to the total number of all possible trophic links (Dunne,
Williams,&Mar tinez,2002),whilethelatterrepresentsthehighest
trophic level occupied by a consumer in their respective food webs
(Digel, Curtsdotter,Riede, Klarner, & Brose,2014). Wecalculated
both using the NetIndices*1.4.4statisticalpackage(Kones,Soetaer t,
van Oevelen, & Owino, 2009).
2.3 | Statistical analysis
We quantified changes in fish assemblages over the two time pe-
riods across flow- modification regimes using the following met-
rics: (a) species richness; (b) proportion of native species; (c) overall
functional diversity; and (d) native functional diversity. We fitted
all metrics and two predictor variables, namely, time period (time)
and flow- modification regimes (treatment), in a list of candidate gen-
eral linear models (glm) making- up an information- theoretic frame-
work (Appendix C) using the lme4*1.1.13statisticalpackage(Bates,
FIGURE1 MapsofSingaporeislandin
both time periods, 1957–1966 (above) and
2006–2010 (below), detailing the location
of our study sites. Flow- modification
regime (i.e., treatment) are represented
by unshaded circles (lentic control), filled
squares (canalised sites) and filled circles
(impoundedsites)respectively.Map
of Singapore in 1966 after Alfred, E. R .
(1966)
|
5
LIEW Et a L.
Maechler,Bolker,& Walker,2015). Fromthis, themodelsbest de-
scribing our data were selected using Akaike’s information criteria
correctedforsmallsamplesizes(AICc; Burnham & Anderson, 2002),
where lower AICc scores indicate greater model parsimony. We cal-
culated AICc scores using the AICcmodavg*2.1.1 statistical package
(Mazerolle,2017).
We ranked all candidate models (and their associated predictor
variables) by comparing their respective Akaike weights (w; Burnham
& Anderson, 2002; Giam & Olden, 2016a), where the Akaike weight
of the i- th model (wi) is
Here, AICc,i represents the AICc value of the i- th model, while
AICc,min represents the AICc value of the most parsimonious model
(i.e., lowest AICc), and AICc,r represents the AIC value of the r-th
model where R is the total number of candidate models. The Akaike
weight (wi) of the i- th model estimates the probabilit y (0–1) that this
model is the best approximating model among the R competing can-
didate models, given the data.
A similar approach was adopted to quantify changes in food web
structure across time and treatment using (a) connectance; and (b)
maximum trophic level as response variables. Again, both metric s
were fitted in candidate glm with combinations of the predictor vari-
ables (Appendix C). Relative model parsimony and predictive ability
were evaluated with AICc and w respectively.
Besides assessing fish assemblage and food web changes be-
tween the two time periods and across flow- modification regime
(at the community level), we investigated probable mechanisms of
change at the species level. To this end, we assessed potential drivers
of species extirpation (since the first time period), as well as trends
in functional changes driven by species establishment (in the second
time period). Our species- level response variables were extinct and
establish. The firs t (i.e., extinct) describes a b inary variabl e where spe-
cies occurring in the first and second time period were ascribed with
azerovaluewhilespeciesoccurringinthe firsttimeperiodbutnot
the secon d were ascribed with the value one. Conversely, our second
species- level response variable, establish, is a binary variable where
species absent in the first time period, but present in the second,
were ascribed with the value one, and all species present in the first
timeperiod were ascribeda zero value,regardless ofwhetherthey
were locally extinct or ex tant in the second time period. Whereas
extinct is indicative of what functional traits predic t for likelihood of
extirpation following flow modification, establish reflects functional
traits most representative of changes in fish communities between
the time periods. Datasets for extinct (n = 104)andestablish (n = 221)
comprised site- specific species occurrences; thus, repeats are pos-
sible for species recorded from multiple study sites. Because mech-
anisms underlying assemblage and/or food web changes may not
be consistent across flow- modification regime, we further divided
dataset s for extinct and establish according to treatment t ype. This
gave us the following: (a) extinctlentic control (n = 20); (b) extinctcanalisation
(n = 32);(c)extinctimpoundment (n = 52); (d) establishlentic control (n = 41);
(e) establishcanalisation (n = 38);and(f)establishimpoundment (n = 142).
The binar y responses extinct and establish were modelled as bi-
nomial responses with logistic link functions and species identity
as random effects to account for repeated species across multiple
study sites in respective sets of candidate generalised linear mixed-
effects models (glmm) (six sets described above). Candidate models
were parameterised using combinations of the following functional
traits: (a) feeding guild (i.e., categorical variable with nine levels: ben-
thic omnivore, detritivore, general predator, herbivore, invertivore,
macroinvertivore, omnivore, pelagic invertivore, pelagic omnivore);
(b) trophic level; (c) maximum length; (d) trophic breadth (i.e., ordinal
variable ranging 1–7 reflecting the number of food items a species’
diet comprises from among the following categories; detritus, plant
matter, pelagicphytoplank ton,periphy ton,zooplankton, macroin-
vertebrates and fishes); (e) habit at preference (i.e., categorical vari-
able with three levels; lentic, lotic or generalist); (f) water column
preference (i.e., categorical variable with three levels; demersal,
benthopelagic and pelagic); (g) body shape (i.e., continuous vari-
able consisting of length–weight ratio parameter a describing body
shape obt ained from Fis hBase) (Froese, Tho rson, & Reyes, 2014);
(h) resilience (i.e., categorical variable derived from a species’ pop-
ulation doubling time with four levels; high, medium, low and very
low); and (i) reproductive strategy (i.e., categorical variable with
ten levels; bubble nest builder, cave brooder, egg scatterer, hard
substrate spawner, soft substrate spawner, mouth brooder, nest
brooder, external brooder, vegetation spawner and live- bearer). In
addition to functional traits, we also tested models parameterised
with fish status (i.e., categorical variable with two levels; native
and alien). Because the species identity random effect was coded
to account for repeat occurrences across multiple study sites, we
compared equivalent general linear models (glm) for lentic control
dataset s (i.e., extinct.lentic control and establish.lentic control) where
this was irrelevant. All candidate glmm/glm were modelled using the
lme4*1.1.13statisticalpackage(Batesetal.,2015).
Wecomparedatotalof31candidatemodelsforeach ofthesix
extinct and establish data set s respectively (Appendix D). To avoid
overfitting, we modelled candidate glmm with a maximum of three
predictor variables. The parameterisation of models with more than
one predictor was broadly guided by the grouping of predictor vari-
ables according to shared relevance to a particular aspect of fish
ecology. For example, feeding guild, trophic level, trophic breadth
and maximum length were grouped because of their relevance to
feeding ecology while habit at preference, water column preference
and maximum length are potential indicators of physical habitat
niches occupied/preferred. Our parameterised models comprised
permutations of variables within these broad groups. Because fish
status (i.e., native or alien) is not a functional trait, the predictor
was not parameterised with other predictors. We evaluated relative
model parsimony and predictive ability with AICc and w respectively.
Statistical analyses were conducted in the R statistical environment
ver.3.4.1(RCoreTeam2017).
w
i=
exp
−1
2
AICc,i−AICc, min
R
r=1exp −1
2
AICc,r−AICc, min
.
6
|
LIEW Et aL.
3 | RESULTS
From the 82 unique fish species encountered over both time periods,
the families Cyprinidae and Cichlidae were most common, with 20
and 11 species respectively. The lat ter is notable bec ause the family
Cichlidae is not native to South- East Asia including Singapore, and
only one cichlid species (i.e., Oreochromis mossambicus) was present
in the first time period. The fish species we recorded were largely
benthopelagic. Egg scattering was the most common reproduc tive
strateg y while general predator, invertivore, macroinvertivore and
omnivore were the dominant feeding guilds.
Werecorded40speciesfromthefirst time periodand63spe-
cies in the second time period (Appendix A). In total, we observed 79
cases of site- sp ecific extirpations. He re, a single species m ay account
for more than one case if found in, and subsequently extirpated
from, more than one site. While there were 25 cases of site- specific
persistence, this comprised only 16 unique species. Conversely, our
data showed 117 cases of site- specific species establishment com-
prising54uniquespecies.
Our Gower’s dissimilarity dendrogram shows less functional
overlap between time periods in canalised and impounded sites
than in the lentic control (Figure 2). However, changes in fish com-
munity metrics were neither generalisable across time nor treatment
(Figure3a). Ourdata suggest that only twoofthefour community
metrics, namely species richness and proportion of native species,
varied with time and/or treatment (Table 2). While species richness
was best predicted by the interactive effects of time and treatment,
propor tion of native species was predicted only by time. There were
no clear trends in both measures of functional diversity (i.e., overall
functional diversity and native functional diversity) across levels in
time and treatment. Our statistical analyses show that while changes
in fish species richness across time periods differed between flow-
modification regimes, the propor tion of native species generally de-
clined over time (Table 2).
Food web structure of our sites across both time periods was as-
sociated with connectance values ranging from 0.119 to 0.286, while
maximumtrophiclevelrangedfrom3.2to4.1(AppendixE).Ourdata
indicate that complexity (i.e., connectance) was neither predicted by
time nor by treatment ( Table 2), but maximum trophic level was best
predicted by the interactive effec ts of both variables (Table 2). This
suggests that changes in maximum trophic level were not consistent
across modification regimes. While lentic control and impounded
sites had higher maximum trophic levels in the second time period,
theoppositewastrueforcanalisedsites(Figure3b).
High rates of extirpation (i.e., 79 cases) and establishment (i.e.,
117 cases) reflect species assemblage turnover between the first
and second time periods. Our data suggest that mechanisms under-
lying species extirpation and establishment- driven functional shifts
differ across flow-modification regimes (Table3). In impounded
rivers, water column preference was an important predictor of ex-
tirpation likelihood, where benthopelagic species were most at risk
(relative to demersal and pelagic species)(Table4).Extirpated spe-
cies were replaced by largely alien taxa. On the other hand, species
extirpation and subsequent establishment in canalised rivers were
predictedneitherbyfunctionaltraitsnorbyfishstatus( Table4).At
the relatively unchanged lentic control, best models for extinct and
establish comprised maximum length (both extinct and establish) and
body shape (only establish). Here, smaller species were more likely to
FIGURE2 Cluster dendrogram of fish assemblages present in different flow regimes of both time periods (i.e., 1957–1966 and 2006–
2010) constructed using species functional traits. Study sites were grouped as pre- and postimpoundment (i.e., predam, postdam), pre- and
postcanalisation (i.e., precanal, postcanal), as well as lentic control at corresponding time periods (i.e., lentic control (~1966), lentic control
(~2010)). Greater heights of node splits (vertical axis) suggest greater differences in functional traits. Lotic and lentic communities are
represented by black and grey bars respectively. Species names and functional traits associated with abbreviations used in the figure are
denoted in Appendix B
|
7
LIEW Et a L.
be extirpated while species occurring in the second time period were
largerandlessvermiformordorsal-ventrallyflat tened(Table4).
4 | DISCUSSION
Our data show that fish assemblages differed bet ween time periods
but the changes varied with flow- modification regime. Qualitatively,
the fish species we encountered can be grouped into two catego-
ries, lotic and lentic. In general, lentic sites (i.e., lentic control, im-
pounded rivers) were associated with a greater suite of functional
groups when compared with lotic sites (i.e., pre- impoundment riv-
ers, precanalisation rivers, postcanalisation rivers). The latter mainly
comprised species typically occurring in small and relatively shal-
low habitats (i.e., left cluster in Figure 2; Baker & Lim, 2012). These
were a mix of native and alien species including the highly success-
ful invader, Poecilia reticulata. Native species belonging to this cat-
egory are largely forest stream specialists. Examples such as Boraras
maculatus, Hemirhamphodon pogonognathus, Silurichthys hasseltii and
Pangio semicincta are not endemic to our sites, but are relatively
restricted in their distribution. Like many forest stream specialists,
thesespeciesoccurinpocketsofforestsinSundaland(i.e.,Malaysia,
Singapore and Thailand) where their habitats are threatened by de-
velopment, agriculture (e.g., oil palm plantations; Giam et al., 2015)
andlogging(Sodhi,Koh,Brook,&Ng,2004).
The broad qualitative lotic–lentic clusters suggest that the im-
pacts of flow modification on fish assemblages are contingent
on whether a lotic–lentic conversion (i.e., impoundment) occurs.
Findings from our quantitative assessments are congruent with
this, as changes in fish species richness over time dif fered between
impounded and ca nalised sites (Table2, Figure3a). This is unsur-
prising considering the influence of habit at complexity on species
richness(Allouche, Kalyuzhny, Moreno-Reuda,Pizarro,&Kadmon,
2012; Loke & Todd, 2016; St. Pierre & Kovalenko, 2014; Stein,
Gerstner, & Kreft, 2014). While impoundment s increase habitat
heterogeneity (e.g., greater range of water columns, more varied
littoral zones), homogenisation of flow and benthos is implicit in
canalisation. Commonly, homogenisation results in lower resource
levels (O’Connor, 1991), fewer refugia (Xavier et al., 2012) and re-
duced flow v ariability (M illidine etal., 2012). T hese outcomes a re
FIGURE3 Indicators of (a) fish assemblage (i.e., species richness, proportion of native species, overall functional diversity, and native
functional diversity) and (b) food web structure (i.e., connect ance, maximum trophic level) at two time periods under different flow-
modification regimes (i.e., treatment). Grey bars represent values recorded in the first time period (i.e., before flow modification), while
white bars represent values measured in the second time period (i.e., post- flow modific ation). Error bars indicate ±1 standard error of the
respective means
8
|
LIEW Et aL.
detrimental to species richness because they represent a net loss of
biologicalniches(Loke&Todd,2016;St.Pierre&Kovalenko,2014).
In addition, increases in species richness after impoundment could
also be attributed to pre- flow- modification communities associ-
ated with low- order stream habitat s similar to our study sites being
species-poortobeginwith(Lotrick,1973).
TABLE2 Details and interpretations of the most parsimonious models describing the relationships between measures of fish assemblage
(i.e., species richness, proportion of native species, overall functional diversity and native functional diversity) and food web proper ties with
treatment type and time
Response variable Most parsimonious model
Akaike weight
(w)Interpretation
Fish species richness (rich)Species richness ~ time*treatment 0.99 Fish species richness differ s between time
periods but changes are not uniform across
flow- modification regimes.
Propor tion of native
species (native)
Propor tion of native species ~ time 0.93 The proportion of native species is different
between time periods across flow- modification
regimes.
Functional diversity (Q)Overall functional diversity ~ 1 0.49 Overall func tional diversity of fish communities
did not dif fer bet ween time periods or
flow- modification regimes.
Native functional diversity
(Qn)
Native functional diversity ~ 1 0.32 Functional diversit y of native species present in
the communities s tudied did not differ between
time periods or flow- modification regimes.
Food web complexity Connectance ~ 1 0.53 Food web complexity in the communities studied
did not dif fer bet ween time periods or
flow- modification regimes.
Maximumtrophiclevel Maximumtrophiclevel~time*treatment 0.35 Maximumtrophicleveldiffersbetweentime
periods but changes are not uniform across
flow- modification regimes.
TABLE3 Summary and interpretations of the most parsimonious models describing: (a) the association between functional traits and
likelihood of species extirpation post- flow modification; and (b) functional changes driven by species establishment post- flow modification
across different modification regimes (i.e., impoundment, canalisation and lentic control)
Response|treatment Most parsimonious model
Akaike weight
(w)Interpretation
extinctlentic control Extinct ~ maximum length 0.22 Species ma ximum length predict s the likelihood
of species extirpation in t he lentic control study
site.
extinctcanalisation Extinct ~ 1 + (1|species) 0.27 Func tional traits do not predict the likelihood of
species extirpation following river canalisation,
after controlling for species identity as a random
effect.
extinctimpoundment Extinct ~ water column
preference + (1|species)
0.43 Species water column preference predict s the
likelihood of species extirpation following river
impoundment, after controlling for species
identit y as a random effect.
establishlentic control Establish ~ shape + maximum length 0.35 Species est ablished in the lentic control study
site in the second time period differs from the
fish assemblage in the first time period in their
bodyshapeandsize,aftercontrollingfor
species identit y as a random effect.
establishcanalisation Establish ~ 1 + (1|species) 0.20 Species newly e stablished following river
canalisation are not func tionally distinct from
fish assemblages occurring in the first time
period, after controlling for species identity as a
random effect.
establishimpoundment Establish ~ status + (1|species) 1.00 Species established following river impoundment
are more likely to be alien, after controlling for
species identit y as a random effect.
|
9
LIEW Et a L.
Unlike species richness, decreases in proportion of native spe-
cies were consistent across flow- modification regimes (Table 2). This
agrees with literature identifying modified water bodies, especially
impoundments, as pathways (sensu Lodge et al., 2006) for biological
invasions(Johnson,Olden,&VanderZanden,2008;Liew,Tanetal.,
2016). Susceptibility to biological invasions are thought to be deter-
mined by levels of propagule pressure (Simberloff, 2009) and biotic
resistance(Liew,Carrascoetal.,2016;Romanuk,Zhou,Valdovinos,
& Marti nez, 2017), both of which a re often elevated in m odified
water bodies. In the specific context of our study sites, ease of public
access (Yeo & Lim, 2011) increases propagule pressure in the form of
releases from the ornamental fish trade. Notable examples include
Acarichthys heckelii, Potamotrygon motoro and Scleropages formosus
(Liew, Tan, Yi, & Yeo, 2014; Ng & Tan, 2010; Ng, Tan, Yeo, & Ng,
2010). At the same time, biotic resistance in modified water bodies
from native communities largely adapted to natural forest stream
habitat s (Baker & Lim, 2012) is unlikely to be significant.
Our study adds to the scarce food web literature documenting
the e f f e c t s o f f lowm o d i f i c a t ion.M a x i m u m foodc h a i n l eng t hha s p r e -
viously been shown to be greater in impounded rivers (Hoeinghaus
etal., 20 08), and our fi ndings agree wi th this (Table2; Figur e3b).
Unlike impounded rivers, canals reflected a decrease in maximum
trophic level and the underlying causes are likely to be analogous
to determinants of species richness discussed earlier. Specifically,
lower resource availability (O’Connor, 1991) in canals limits trophic
levels because some energy is lost to entropy with each trophic
transfer(Takimoto&Post,2013).
We did not expect food web complexity to remain statistically
invariant across time periods, especially at impounded sites. This is
because a wider range of biological niches resulting from increased
habitat heterogeneity can also create conditions that are more
conducive to adaptive prey- switching commonly linked to greater
food web complexity (Uchida, Drossel, & Brose, 20 07). In canalised
sites, an influx of non- native species with generalist diets could fea-
sibly increase food web complexity via the same mechanism (i.e.,
adaptive prey- switching). Considering that resource availability is
the other known determinant of food web complexity (Liew et al.,
2018), a lack of variation in food web complexity at our sites may re-
flect no net change in per capita resource levels between time peri-
ods, masking the influence of other environmental changes resulting
from flow modification.
Shifts in fish assemblages were reflected by species turnovers
(via extinction and establishment), and again, our observations sup-
port the cogency of lotic- to- lentic conversions in dictating direction-
ality in f unctional cha nge associated with f low modificati on. Our data
from impounded rivers are congruent with global patterns of fish
assemblage change (Liew, Tan et al., 2016) where the replacement of
extirpatedriverinespecies byalientaxa followsdamming (Tables3
and 4). The im portance of w ater column prefer ence in predic ting
likelihood of extirpation following river impoundment suggests that
assemblage change in impoundments is a function of physical hab-
itat changes in that impoundments are generally larger and deeper
than natural streams. Crucially, our data show that the influence of
lotic- to- lentic conversions on fish communities persists long after
impoundment (e.g., at the lentic control site). Here, changes in fish
assemblages favouring larger, more robust fish species are also con-
sistent with shifts in available physical niches. Our findings reinforce
the hypothesis that impoundments shape the functional assemblage
of associated communities (Olden, Poff, & Bestgen, 20 06).
The apparent stochasticit y in assemblage change at canalised
sites suggests that when flow modification does not involve a state
transition (e.g., when stream is canalised rather than impounded,
TABLE4 Details of the most parsimonious models describing likelihood of extirpation and functional changes in fish communities across
flow- modification regimes. For extinct, larger coefficients in categorical predictors suggest higher likelihood of extirpation in species
associated with the trait level, relative to species associated with trait levels of lower coefficient values. Conversely, if the predictor is
continuous, positive coefficients suggest greater likelihood of extirpation in species attributed with greater trait values (e.g., greater
maximum length) and vice versa. For establish, positive coefficients in categorical predictors suggest greater likelihood of occurring in the
second time period if species are associated with the trait level, relative to species associated with trait levels of lower coefficient values. If
thepredictoriscontinuous,positivecoefficientssuggestgreateroddsofhighertraitvaluesoccurringinthesecondtimeperiod.Effectsizes
presented in the table represent odds ratios
Response|treatment Model Predictor variables
Levels (categorical
predictors) Coefficient Effect size
extinctlentic control Extinct ~ maximum length Maximumlength NA −0.01 1.00
extinctcanalisation Extinct ~ 1 + (1|species) NA NA NA NA
extinctimpoundment Extinct ~ water column
preference + (1|species)
Water column
preference
Pelagic −35.48 3.90×10−16
Demersal −0.9 2 0.40
Benthopelagic 1.62 5.05
establishlentic control Establish ~ shape + maximum length Body shape NA 114.0 0 3. 23×1049
Maximumlength NA 0.01 1.02
establishcanalisation Establish ~ 1 + (1|species) NA NA NA NA
establishimpoundment Establish ~ status + (1|species) Status Native −9.12 1.10×10−4
Alien 10.10 2.43×104
10
|
LIEW Et aL.
the habitat remains lotic), population density/distribution may be
better predictors of extinction probabilities (Giam, Ng, Lok, & Ng,
2011). Conversely, community assembly (sensu Giam & Olden,
2016b) in these urban water bodies may be driven by anthropogenic
influences, with levels and frequency of species introductions (i.e.,
propagule pressure; Simberloff, 2009) a likely candidate. This con-
tradicts findings linking species traits to establishment likelihood
(Vila-Gisper t, Alc araz, & Garcia-Berthou, 20 05), but is not without
precedent. Recent investigations of invasion success suggest that
human use of alien species (e.g., for consumption) can be the ulti-
mate predictor of establishment likelihood, sometimes confound-
ingmoreproximal,traits-basedassessments ofinvasiveness(Zeng,
Chong, Grey, Lodge, & Yeo, 2015).
Despite ef forts to minimise comparative biases that may stem
from differences in sampling methodology, there are potential cave-
ats to our findings. Primarily, the detection probability of rare taxa
may not be equal between data sources and could be higher in the
first time period given greater temporal coverage. In the context of
our findings, this means that species we listed as locally ex tinct in the
second time period may instead have reduced population densities.
For example, water column preference could be more conservatively
interpreted as being predictive of vulnerability to population den-
sity loss and/or range restriction in impounded rivers. In the event
that extirpation was wrongly assumed, the scarcity of misclassified
species may nevertheless suggest functional extirpation (sensu
Barnosky et al., 2011) with the potential for complete loss over time,
should modified conditions persist.
Our study shows that flow modification influences freshwa-
ter fish assemblages and food webs, albeit in contrasting ways.
Mechanistically, differences in the directionality (or lack thereof)
in postmodification shif ts appear to be contingent on whether a
lotic- to- lentic flow conversion was effected. The nuances our data
described in this paper suggest that studies conducted at multiple
scales (e.g., at community and species levels) and approaches (e.g.,
assessing species assemblage and food webs) may be necessary to
fully understand the varied effects of anthropogenic exploitation of
natural resources.
ACKNOWLEDGEMENTS
We gratefully acknowledge two anonymous reviewers whose com-
ments and suggestions helped improve earlier versions of this manu-
script. We thank the Public Utilities Board of Singapore (National
University of Singapore Grant No. R-154- 000-619-490) and the
National Research Foundation and the Economic Development
Board (SPORE, COY- 15- EWI- RCFSA /N197- 1) for financial support.
We also thank members of the National University of Singapore
(NUS) Reser voir Biodiversit y team for contribution to fish assem-
blage data.
ORCID
Jia Huan Liew http://orcid.org/0000-0002-7649-0398
REFERENCES
Alfred, E. R. (1966). The fresh–water fishes of Singapore. Zoologische
Verhandelingen, 78, 5–67.
Allan, J. D.,Abell,R .,Hogan, Z., Revenga, C., Taylor,B.W., Welcomme,
R. L., & Winemiller, K. O. (2005). Over fishing of inland waters.
BioScience, 55, 1041–1051. http s://doi.org /10.1641/00 06-3568(20
05)055[1041:OOIW]2.0.CO;2
Allouche,O., Kalyuzhny,M.,Moreno-Reuda,G.,Pizarro,M.,& Kadmon,
R. (2012). Area- heterogeneity tradeoff and the diversit y of ecolog-
ical communities. Proceedings of the National Academy of Sciences
of the United States of America, 109, 17495–17500. https://doi.
org /10.1073/pnas.1208652109
Baker, N., & Lim, K. K. P. (2012). Wild animals of Singapore. Singapore:
Draco Publishing and Distribution.
Barnos ky, A. D., Matzke , N., Tomiya, S., Wogan , G. O. U., Swart z, B.,
Quental, T. B., … Ferrer, E. A. (2011). Has the Earth’s sixth mass ex-
tinction already arrived? Nature, 471 , 51–5 7. ht t p s : //d oi . o r g /1 0. 1 03 8 /
nature 09678
Bates, D., M aechler, M., Bolke r, B., & Walker, S. (2015) . Fitting line ar
mixed-effects models using lme4. Journal of Statistical Software,
67(1),1–48.
Botta-Dukat , Z. (20 05). Rao’squadratic entropy as a measure of func-
tional diversit y based on multiple trait s. Journal of Vegetation Science,
16,533–540.https ://doi.or g/10 .1111/j.165 4-1103.2005.tb 02393. x
Boulton,A.J.,Boyero,L.,Covich,A.P.,Dobson,M.,Lake,S.,&Pearson,
R. G. (2008). Are tropical streams ecologically different from tem-
perate streams? In D. Dudgeon (Ed.), Tropical stream ecology (pp.
257–284). Hong Kong: Academic Press. https://doi.org /10.1016/
B978-012088449- 0.50011-X
Boyero,L.,Ramirez,A.,Dudgeon ,D.,&Pearson,R.G.(2009).Aretropical
streams really different? Journal of the North American Benthological
Society, 28,397–403.https://doi.org /10.1899/08-146.1
Burnham, K. P., & Ander son, D. R. (2002). Model selection and multimodel
inference: A practical information-theoretic approach. New York, NY:
SpringerScience&BusinessMedia.
Chao, A., & Jost, L. (2012). Coverage- based rarefaction and extrap-
olation: Standardizing samples by completeness rather than size.
Ecology, 93(1 2), 253 3–2547.h tt ps: //doi.o rg /10.1890/11-1952.1
Correa, C., Bravo, A. P., & Hendr y, A. P. (2012). Reciprocal trophic
niche shifts in native and invasive fish: Salmonids and galaxiids in
Patagonian lakes. Freshwater Biology, 57, 1769–1781. https://doi.
org /10.1111/j .1365-2427.2 012 .02837.x
Didham,R.K.,Tylianakis, J.M.,Gemmell,N. J., Rand,T.A .,&Ewers,R .
M.(2007).Interactiveeffec tsofhabitatmodificationandspeciesin-
vasion on native species decline. Trends in Ecology & Evolution, 22,
489–496.https://doi.org/10.1016/j.tree.2007.07.001
Digel, C., Curtsdotter, A., Riede, J., Klarner, B., & Brose, U. (2014).
Unravelling the complex structure of forest soil food webs: Higher
omnivory and more trophic levels. Oikos, 123, 1157–1172. https://
doi .org/10.1111/oik.0 0865
Dudgeo n, D. (2014). Threats to fr eshwater biodiv ersity in a chan ging
world. In B . Freedman (Ed.), Global Environmental Change(pp.243–
253).Dordrecht,theNetherlands:SpringerNetherlands.
Duedoro,S., Box, A.,Vazquez-Luis,M., & Arroyo, N.L . (2014).Benthic
community responses to macroalgae invasions in seagrass beds:
Diversity, isotopic niche and food web structure at community
level. Estuarine, Coastal and Shelf Science, 142, 12–22. https://doi.
org/10.1016/j.ecss.2014.03.006
Dunne,J.A.,Williams,R.J., &Martinez,N. D.(2002).Foodwebstructure
andnetwork theory:Theroleofconnectance andsize. Proceedings of
the National Academy of Sciences of the United States of America, 99,
12917–12922.https://doi.org/10.1073/pnas.192407699
Edelman, A., Gelding, A., Konovalov, E., McCominskie, R., Penny, A.,
Nicholas, R., … Turto n, S. (2014). State of the tropics 2014 repor t.
Cairns, Qld, Australia: James Cook University.
|
11
LIEW Et a L.
Froese, R ., & Pauly, D. (2011). Fishbase, Fishbase.
Froese, R., Thor son, J.T.,&Reyes,R.B. Jr(2014).ABayesian approach
for estimating length- weight relationships in fishes. Journa l of Applied
Ichthyology, 30, 78–85. https://doi.org/10.1111/jai.12299
Giam, X., Hadiat y, R., Tan, H. H., Parenti, L. R., Wowor, D., Sauri, S., …
Wilcove,D.S.(2015).Mitigatingtheimpactofoilpalmmonoculture
on freshwater fishes in Southeast A sia. Conservation Biology, 29,
1357–1367.ht tp s://doi .org/10.1111/cobi.124 83
Giam, X.,Koh, L. P.,Tan, H. H.,Miettinen, J., Tan,H. T.W., & Ng, P.K.
L. (2012). G lobal ex tinctions of freshwater f ishes follow peatland
conversion in Sundaland. Frontiers in Ecology and the Environment, 10,
465–470.https://doi.org/10.1890/110182
Giam, X., Ng, T. H., Lok , A. F. S. L., & Ng, H. H. (2011). Local
geographic range predic ts freshwater fish extinctions in
Singapore. Journal of Applied Ecology, 48, 356–363. https://doi.
org/10.1111/j.1365-2664.2010.01953.x
Giam, X., & Olden, J. D. (2016a). Quantifying variable importance in a
multimodel inference framework. Methods in Ecology an d Evolution, 7,
388–397.https://doi.org/10.1111/2041-210X.12492
Giam, X., & Olden, J. D. (2016b). Environment and predation govern
fish community assembly in temperate streams. Global Ecolog y and
Biogeography, 25,1194–1205.https://doi.org/10.1111/geb.12475
Gibbs, M.A .,Shields,J.H.,Lock,D.W.,Talmadge,K .M., &Farrell,T.M.
(200 8). Reproduct ion in an invasive e xotic catfis h Pterygoplichthys dis-
junctivus in Volusia Blue Spring , Florida, U.S. A. Journal of Fish Biolog y,
73,1562–1572.https://doi.org /10.1111/j.1095-8649.20 08.02031.x
Goncalves, J. F. Jr, Graca, M. A. S., & Callisto, M. (2006). Leaf-
litter breakdown in 3 streams in temperate, Mediterranean,
and tropical Cer rado climates. Journal of the North
American Benthological Society, 25, 344–355. https://doi.
org/10.1899/0887-3593(2006)25[344:LBISIT]2.0.CO;2
Gopal, B., & Ghosh, D. (2009). Lakes and reser voirs of A sia. In G.
E. Likens (Ed.), Encyclopedia of inland waters (pp. 501–512).
Amsterdam, the Netherlands: Elsevier. https://doi.org/10.1016/
B978-012370626-3.00032-6
Gower, J. C. (1971). A general c oefficie nt of similarit y and some of it s prop-
erties. Biometrics, 27,857–871.https://doi.org/10.23 07/2528823
Hoeinghaus, D., Winemiller, K. O., & Agostino, A . (200 8).
Hydrogeomorphology and river impoundment affect food- chain
length of diverse Neotropical food webs. Oikos, 117, 984–995.
https ://doi.or g/10 .1111/j.0 030 -1299.20 08 .16459.x
Hsieh, T. C., Ma, K. H., & Chao, A . (2016). An R package for rar-
efaction and ex trapolation of species diversit y (Hill numbers).
Methods in Ecology and Evolution, 7, 1451–1456. h tt ps ://doi .
org /10.1111/20 41-210X .12613
Johnson,P.T.,Olden, J.D.,& Vander Zanden,M.J. (20 08). Daminvad-
ers: Impoundments facilitate biologic al invasions into freshwaters.
Frontiers in Ecology and the Environment, 6, 357–363. https://doi.
org/10.189 0/070156
Kones, J. K ., Soet aert, K., van Oevelen, D., & Owino, O. (2009). Are net-
work indi ces robust in dicators of fo od web funct ioning? A Monte
Carlo approach. Ecological Modelling, 220, 370–382. https://doi.
org/10.1016/j.ecolmodel.2008.10.012
Liew, J. H., Carrasco, L. R., Tan, H. H., & Yeo, D. C. J. (2016). Native rich-
ness and species level trophic traits predict est ablishment of alien
freshwater fish. Biological Invasions, 18, 3495–3512. https://doi.
org /10.1007/s10530-016 -1241-z
Liew,J.H.,Jardine,T.D.,Li m,B .H.R .,Kwik,J.T.B.,Tan,H.H.,Kho,Z .Y.,
& Yeo, D. C. J. (2018). Bot tom- up influences on freshwater food web
struc ture in the tropics support the ‘environmental filtering’ hypoth-
esis. Limnology & Oceanography.ht tp s://doi.org/10.1002/l no.10 813
Liew, J. H., Tan, H. H ., & Yeo, D. C. J. (2016). Damme d rivers: Im poundmen ts
facilitate fish invasions. Freshwater Biology, 61, 1421–1429. h ttp s://
doi .org/10.1111/fw b.12781
Liew, J. H., Tan, H. H., Yi, Y.,& Yeo, D. C. J. (2014). Ecology and origin
of the introduced cichlid Acarichthys heckelii in Singapore’s fresh
waters- first instance of establishment. Environmental Biology of
Fishes, 97,1109–1118.https://doi.org /10.10 07/s10641-013-02 01-z
Liu, F., Wang, J., & Cao, W. (2012). Long- term changes in fish assem-
blage following the impoundments of the Three Gorges Reservoir in
Heijiang,aprotectedreachoftheupperYangtzeriver.Knowledge and
Management of Aquatic Ecosystems, 407, 2–1 7.
Lodge, D. M., Williams, S., MacIsaac, H. J., Hayes, K. R., Leung,
B., Reichard, S., … McMichael, A. (2006). Biological invasions:
Recommendations for U.S. policy and management. Ecological
Applications, 16,2035–2054.https://doi.org/10.1890/1051-0761(20
06)016[2035:BIRFUP]2.0.CO;2
Loke, L. H . L., & Todd, P. A. (2016). Str uctural complexity and component
type increase intertidal biodiversity independently of area. Ecology,
97,383–393.https://doi.org/10.1890/15-0257.1
Lotrick , V.A . (1973).G rowth, production,and community composition
of fishes inhabiting a firs t- , second- , and third order s tream of east-
ern Kentucky. Ecological Monographs, 43(3), 377–397. https://doi.
org /10. 2307/1942347
Lytle, D. A ., & Poff, N. L. (2 004). Adap tation to natu ral flow regim es.
Trends in Ecology & Evolution, 19,94–100. h ttps://doi. org /10.1016/ j.
tree.2003.10.002
Magurran, A.E., & Queiroz, H. (2010).Evaluating tropic al biodiversit y:
Do we need a more refined approach. Biotropica, 42, 537–539.
https://doi.org/10.1111/j.1744-7429.2010.00670.x
Mazerolle,M. J. (2017). AICcmodavg: Model selection and multimodel in-
ference based on (Q)AIC(c). R package version 2 .1-1. Retrieved from
https://cran.r-project.org/package=AICcmodavg
McNew,L.B.,&Handel,C.M.(2015).Evaluatingspeciesrichness:Biased
ecological inference results from spatial heterogeneity in detection
probabilities. Ecological Applications, 25(6), 1669–1680. https://doi.
org /10.1890/14 -1248.1
Melche r,A . H., Ouedr aogo, R., & S chmutz, S . (2012). Spatia l and sea-
sonal fish community pat terns in impacted and protected semi- arid
rivers of Burkina Faso. Ecological Engineering, 48, 117–129. https://
doi.org/10.1016/j.ecoleng.2011.07.012
Mercado-Silva, N ., Helmus, M. R.,&Vander Zan den, M. J.(2008).The
effec ts of impoundment and non- native species on a river food
web in Mexi co’sc entral plate au. River Research & Applications, 25,
1090–1108.
Millennium Ecosys tem Assessment (2005). Ecosystems and human
well-being: Synthesis. Washington, DC: Island Press.
Millidine, K. J., Malcolm, I. A ., Gibbins, C . N., Fr yer,R. J., & Youngson,
A. F. (2012). The influence of c analis ation on juvenile salmonid hab-
itat. Ecological Indicators, 23, 262–273. ht tps://doi.org/10.1016/j.
ecolind.2012.04.004
Ng, H. H., & Tan, H. H. (2010). An annotated checklist of the non-
native freshwater fish species. COSMOS, 6, 95–116. https://doi.
org/10.1142/S0219607710000504
Ng, H. H. , Tan, H. H., Yeo, D. C . J., & Ng, P. K. L. (2010). Stinge rs in a strange
land: South American freshwater s tingrays (Potamotrygonidae)
in Singapore. Biological Invasions, 12, 2385–2388. https://doi.
org /10.1007/s10530-0 09-9663-5
O’Connor, N. A. (1991). The effects of habitat complexity on the mac-
roinvertebrates colonising wood substrates in a lowland stream.
Oecologia, 85,504–512.https://doi.org/10.1007/BF00323762
Olden, J. D., Poff, N. L ., & Bestgen, K. R. (20 06). Life- his tory strate-
gies predict fish extirpations in the Colorado River Basin . Ecological
Monographs, 76(1),25–4 0.https://doi.org/10.1890/05-0330
Penczak,T.,Agostinho,A.A.,Gomes,L .C.,&Latini,J.D.(2009).Impacts
of a reser voir on fish assemblages of small tributaries of the Cor umba
river,Brazil.River Research and Applications, 25,1013–1024. https: //
doi.org/10.10 02/rra .120 0
12
|
LIEW Et aL.
Quinn, J.W.,&Kwak, T.J.(2013).Fishassemblage changesinanOzark
River after impoundment: A long- term perspective. Transactions of
the American Fisheries Society, 132, 110–119.
R Core Team (2017). A language and environment for statistical computing.
Vienna, Austria: R Foundation for Statistical Computing.
Rolls, R. J., Growns, I. O., Khan, T. A., Wilson, G. G., Ellison, T. L., Prior,
A., & Waring, C.C .(2013).Fish recruitment in riverswith modified
discharge depends on the interacting effects of flow and thermal re-
gimes. Freshwater Biology, 58, 1 804 –1819.h ttp s://doi .org /10.1111/
fwb.12169
Romanuk , T.N ., Zhou, Y., Valdovinos , F. S., & Mar tinez, N. D. (2017) .
Robustness trade- offs in model food webs: Invasion probabilit y
decreases while invasion consequences increases with connec-
tance. Advances in Ecological Research, 56 , 263–291. https://doi.
org/10.1016/bs.aecr.2016.11.001
Simberloff, D. (2009). The role of propagule pressure in biological inva-
sions. Annual Review of Ecology, Evolution, and Systematics, 40, 81–
102.https://doi.org/10.1146/annurev.ecolsys.110308.120304
Simoes, N. R., Nunes, A . H., Dias, J. D., Lansac-Toha, F. A., Velho, L. F.
M., & Bonecker, C. C . (2015). Impact of reser voirs on zooplankton
diversity and implications for the conser vation of natural aquatic
environments. Hydrobiologia, 758, 3–17. htt ps://doi.org /10.10 07/
s10750-015-2260-y
Singapore Open Data License (2018). https://data.gov.sg/
open-data-licence.
Smith, V. H., Joye, S. B., & Howarth, R. W. (2006). Eutrophication of
freshwater and marine ecosystems. Limnology and Oceanography, 51,
351–355.https://doi.org/10.4319/lo.2006.51.1_part_2.0351
Sodhi,N.S.,Koh,L.P.,Brook,B.W.,&Ng,P.K.L.(2004).Sout heastAsian
biodiversity: And impending disaster. Trends in Ecology and Evolution,
19(12),654–660.https://doi.org/10.1016/j.tree.2004.09.006
St.Pierre,J. I., & Kovalenko,K .E.(2014).Effectsof habitat complexity
attributes on sp ecies richness. Ecosphere, 5, 1–10 .
Stein, A., Gerstner,K., &Kreft,H .(2014).Environmentalheterogeneity
as a universal driver of species richness across taxa, biomes, and spa-
tial scales. Ecology Letters, 17 (7), 866–880. https://doi.org/10.1111/
ele.12277
Strayer, D. L., & Dudgeon, D. (2010). Freshwater biodiversit y con-
servation: Recent progress and future challenges. Journal of the
North American Benthological Society, 29, 344–358. https://doi.
org /10.1899/08-171.1
Takimoto, R. M ., & Post, D. M. ( 2013). Environm ental dete rminant s of
food- chain length: A meta- analysis. Ecological Research, 28(5), 675–
681.https://doi.org/10.1007/s11284-012-0943-7
Thomps on, R. M., Ul rich, B., Du nne, J. A., Ha ll, R. O. J., Hlad yz, S.,
K i t ch i n g , R . L . , … Ty l i a n a k i s ,J . M . ( 2 0 1 2) . F o o d w e bs : R e c o n c i l in g t h e
struc ture and function of biodiversit y. Trends in Ecology & Evol ution,
27(12), 689–697. https://doi.org/10.1016/j.tree.2012.08.005
Tockner,K.,Pusch,M.T.,Borchardt,D.,&Lorang,M.S.(2010).Multiple
stressors in coupled river- floodplain ecosystems. Freshwater Biology,
55,135–151.https://doi.org/10.1111/j.1365-2427.2009.02371.x
Uchida, S ., Drossel, B., & Brose, U. (20 07). The s tructure of food webs
with adaptive behaviour. Ecological Modelling, 206,263–276.https://
doi.org/10.1016/j.ecolmodel.2007.03.035
Van Cappe llen, P., & Maavara, T. (2016). Rive rs in the Anthr opocene:
Global scale modifications of riverine nutrient fluxes by damming.
Ecohydrolog y & Hydrobiology, 16, 10 6–111. https://doi.or g/10.1016/j.
ecohyd.2016.04.0 01
Vaughn, C. (2010). Biodiversit y losses and ecosystem function in fresh-
waters: Emerging conclusions and research directions. BioScience,
60,25–35.https://doi.org/10.1525/bio.2010.60.1.7
Vila-G ispert, A ., Alcara z, C., & Garcia- Berthou, E . (2005). Life -histo ry
traits of invasive fish in small Mediterranean streams. Biological
Invasions, 7,107.https://doi.org/10.1007/s10530-004-9640-y
Vilmi, A ., Karjalainen, S. M., Landeiro, V. L., & Heino, J. (2015).
Freshwater diatoms as environmental indicators: Evaluating the
effec ts of eutrophication using species morphology and bio-
logical indices. Environmental Monitoring and Assessment, 2015,
4485–4487.
Vitule, J. R. S., S kora, F., & Abilhoa, V. (2012). Homogenization
of freshwater fish faunas after the elimination of a natu-
ral barrier. Diversity and Distributions, 18, 111–120. https ://doi.
org /10.1111/j .1472-4642. 2011.0 0821.x
Xavier,J.H. A.,Cordeiro, C. A. M.M.,Tenorio,G.D.,Diniz,A .F.,Paulo
Junior, E. P. N. P., Rosa, R. S., & Rosa, I. L . (2012). Fish assemblage
of the Mamanguape Environmental Protection Area, NE Brazil:
Abundance, composition and microhabitat availability along the
mangrove- reef gradient. Neotropical Ichthyology, 10, 109–122.
htt ps://doi.org/10.159 0/S1679-622520120 00100 011
Yeo, D. C. J., & Lim, K. K. P. (2011). Freshwater ecosystems. In P. K. L. Ng,
R. T. Corlet t, & H. T. W. Tan (Eds.), Singapore biodiversity(pp.52–63).
Singapore:EditionsDidierMillet.
Zeng, Y.,Chong, K.Y., Grey, E. K., Lodge,D.M.,&Yeo, D. C. J. (2015).
Disregarding human pre- introduction selection can confound inva-
sive cray fish risk assessments. Biological Invasions, 17(8),2373–2385.
https://doi.org/10.1007/s10530-015-0881-8
Zhang,X. ,G ao ,X ., Wang,J. ,&Ca o, W.(201 5) .E xtinc ti on ri skan dconser-
vationpriorityanalysesfor64endemic fishesintheUpperYang tze
River, China. Environmental Biology of Fishes, 98, 261–272. https://doi.
org /10.1007/s10641-014- 0257-4
SUPPORTING INFORMATION
Additional supporting information may be found online in the
Suppor ting Information section at the end of the article.
How to cite this article: Liew JH, Giam X , Clews E, et al.
Contrasting changes in freshwater fish assemblages and
food webs follow modification of tropical waterways. Ecol
Freshw Fish. 2018;00:1–12. ht tp s://doi.o rg /10.1111/
ef f.12419
