Comparison of the Abiotic Preferences of Macroinvertebrates in Tropical River Basins

16 Pages·2014·1.06 MB·English

Comparison of the Abiotic Preferences of
Macroinvertebrates in Tropical River Basins
Gert Everaert1,2*, Jan De Neve3, Pieter Boets1, Luis Dominguez-Granda4, Seid Tiku Mereta5,
Argaw Ambelu5, Thu Huong Hoang6, Peter L. M. Goethals1, Olivier Thas3,7
1Aquatic Ecology Research Unit,DepartmentAppliedEcology and EnvironmentalBiology, Ghent University,Ghent,Belgium, 2EnvironmentalToxicology Research
Group, Department Applied Ecology and Environmental Biology, Ghent University, Ghent, Belgium, 3Department of Mathematical Modelling, Statistics and
Bioinformatics,GhentUniversity,Ghent,Belgium,4DepartmentofChemicalandEnvironmentalSciences,EscuelaSuperiorPolite´cnicadelLitoral(ESPOL),Guayaquil,
Ecuador,5DepartmentofEnvironmentalHealthScienceandTechnology,JimmaUniversity,Jimma,Ethiopia,6SchoolofEnvironmentalScienceandTechnology,Hanoi
UniversityofScienceandTechnology,Hanoi,Vietnam,7NationalInstituteforAppliedStatisticsResearchAustralia(NIASRA),SchoolofMathematicsandAppliedStatistics,
UniversityofWollongong,Wollongong,Australia
Abstract
We assessed and compared abiotic preferences of aquatic macroinvertebrates in three river basins located in Ecuador,
Ethiopia and Vietnam. Upon using logistic regression models we analyzed the relationship between the probability of
occurrence of five macroinvertebrate families, ranging from pollution tolerant to pollution sensitive, (Chironomidae,
Baetidae, Hydroptilidae, Libellulidae and Leptophlebiidae) and physical-chemical water quality conditions. Within the
investigatedphysical-chemicalranges,nineoutoftwenty-fiveinteractioneffectsweresignificant.Ouranalysessuggested
river basin dependent associations between the macroinvertebrate families and the corresponding physical-chemical
conditions.Itwasfoundthatpollutiontolerantfamiliesshowednoclearabioticpreferenceandoccurredatmostsampling
locations, i.e. Chironomidae were present in 91%, 84% and 93% of the samples taken in Ecuador, Ethiopia and Vietnam.
Pollutionsensitivefamilieswerestronglyassociatedwithdissolvedoxygenandstreamvelocity,e.g.Leptophlebiidaewere
onlypresentin48%,2%and18%ofthesamplesinEcuador,EthiopiaandVietnam.Despitesomelimitationsinthestudy
design,weconcludedthatassociationsbetweenmacroinvertebratesandabioticconditionscanberiverbasin-specificand
henceare notautomaticallytransferable across river basins inthe tropics.
Citation:EveraertG,DeNeveJ,BoetsP,Dominguez-GrandaL,MeretaST,etal.(2014)ComparisonoftheAbioticPreferencesofMacroinvertebratesinTropical
RiverBasins.PLoSONE9(10):e108898.doi:10.1371/journal.pone.0108898
Editor:SyuheiBan,UniversityofShigaPrefecture,Japan
ReceivedMarch19,2014;AcceptedAugust3,2014;PublishedOctober3,2014
Copyright:(cid:2)2014Everaertetal.Thisisanopen-accessarticledistributedunderthetermsoftheCreativeCommonsAttributionLicense,whichpermits
unrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalauthorandsourcearecredited.
DataAvailability:Theauthorsconfirmthatalldataunderlyingthefindingsarefullyavailablewithoutrestriction.DatahavebeendepositedatDryadandare
availablewiththeDOI:doi:10.5061/dryad.20860.
Funding:PieterBoetsisapostdoctoralresearchfellowfromtheResearchFoundationFlanders(FWO-http://www.fwo.be/en/).LuisDominguez-Grandareceived
financial support of the VLIR ESPOL IUC program in Ecuador and SENACYT (http://www.vliruos.be/en/project-funding/programdetail/institutional-university-
cooperation_3948/andhttp://www.educacionsuperior.gob.ec/).ArgawAmbeluwasarecipientofanICP-PhDscholarshipfromVLIR-UOS(http://www.vliruos.be/
en/countries/countrydetail/ethiopia_3854/).SeidTikuMeretawasarecipientofaIUC-PhDscholarshipfromVLIR-UOS(http://www.iucju.ugent.be/).ThuHuong
HoangreceivedfinancialaidfromtheBelgianTechnicalCooperation(BTC-http://www.btcctb.org/en/countries/vietnam).MarieAnneEurieForioandMinar
NaomiDamanikAmbaritacooperatedwithfinancialsupportoftheVLIREcuadorBiodiversityNetwork(http://www.vliruos.be/en/ongoing-projects/overview-of-
ongoing-projects/network-cooperation/network-biodev-ecuador/).Thefundershadnoroleinstudydesign,datacollectionandanalysis,decisiontopublish,or
preparationofthemanuscript.
CompetingInterests:Theauthorshavedeclaredthatnocompetinginterestsexist.
*Email:[email protected]
Introduction international biotic index. In various regions biotic indices have
beenappliedusingaderivateoftheBMWP,includingtheIberian
Benthic macroinvertebrates have often been used for water BMWP[9]andtheSouthAfricanScoringSystem(SASS,[5]).In
qualitymonitoringandassessment[1,2].Theyaredirectmeasures tropical countries, water quality assessment has been primarily
ofstreamconditions,integratehumanandnaturalstressorsovera performed based on physical-chemical water quality measure-
long period of time and reflect the quality of their surroundings ments [10]. However, in recent years, macroinvertebrate-based
[3].Benthicmacroinvertebratescanbeusedasbio-indicatorssince water quality assessments have also been conducted in tropical
different macroinvertebrate taxa have different tolerances to countries [11].
pollution [4]. Therefore, benthic macroinvertebrates are suitable Most macroinvertebrate-based water quality assessment meth-
for assessing the ecological state of aquatic ecosystems. Macroin- ods have been developed in temperate climate regions where
vertebrate-based water quality assessment methods have increas- relationshipsbetweenenvironmentalvariablesandtheoccurrence
ingly been applied innational monitoring campaigns (e.g., [5,6]). of macroinvertebrate taxa are well documented. Due to the
In the United Kingdom, macroinvertebrate-based water quality infancy of the macroinvertebrate-based monitoring and assess-
assessmenthasbeenappliedsince1970bymeansoftheChandler ment in the tropics [12], the ecological water quality in tropical
Score [7].In lateryears,thebiological monitoring workingparty countries was often assessed based on indices constructed for
(BMWP; [8]) was developed and accepted as a standard temperate climate regions [13,14]. However, it has been shown
PLOSONE | www.plosone.org 1 October2014 | Volume 9 | Issue 10 | e108898
AbioticPreferencesofAquaticMacroinvertebratesintheTropics
thatapplyingecologicalindicesandhabitatsuitabilitymodelsfrom specific permission was needed for our sampling activities nor
temperate regions to the tropics can lead to uncertain ecological locations since we were only interested in macroinvertebrates.
valuations [15,16]. In this perspective, river basin scale habitat Sampling did not involve endangeredor protected species.
suitabilitymodelshaverecentlybeendevelopedforthetropics,e.g.
inEcuador[17],Ethiopia[18,19]andVietnam[20,21].Inspiteof Data exploration
this progress, limited knowledge is available about the habitat Boxplots were used for data exploration. The first series of
preferencesofmacroinvertebratesinriversinthetropics[22].For boxplots visualized the seasonality of the physical-chemical
instance, it is poorly understood whether associations between variables per river basin. The second series of boxplots summa-
macroinvertebrate taxa and environmental conditions vary rizedforeachmacroinvertebratefamilyandperphysical-chemical
betweenriverbasinsandcontinents.Assuch,thequestionremains variableandriverbasintheconditionsunderwhichthetaxonwas
whether the relations obtained between the macroinvertebrate presentandabsent.BoxplotswereconstructedwiththeRsoftware
taxa and the physical-chemical water quality conditions are only [23].
valid in the river basin where they have been developed. Theunivariateassociationsbetweenexplanatoryvariableswere
Therefore, it is questioned whether the abiotic preferences of assessedwithpairwiseSpearman’srankcorrelations,whichisoften
macroinvertebratesandtheirresponsestoenvironmentalpollution usedin ecology duetoitsnonparametric nature [24].
are transferable across riverbasins.
The aim of this study was to assess if abiotic preferences of Logistic regression model
aquatic macroinvertebrates differed between three river basins in Theoccurrenceoffivemacroinvertebratefamilieswasrelatedto
the tropics. We investigated associations between physical- thephysical-chemicalwaterqualityconditionsusingaregression-
chemical variables and macroinvertebrate occurrences in one basedmodelandtheserelationshipswerecomparedbetweenthree
Ecuadorian, oneEthiopian andoneVietnamese riverbasin. Five river basins situated in the tropics. Logistic regression models
representative macroinvertebrate taxa, ranging from pollution (LRMs) were used to infer relationships between occurrences of
tolerant to pollution sensitive were selected and their preferences five aquatic macroinvertebrate families and environmental data.
towards environmental conditions were compared using regres- LRMs have been frequently used to model the presence or
sion-based ecological models. absence of a species in relation to environmental variables [25–
27].Fivelogisticregressionmodels(LRMs)wereconstructed,one
Materials and Methods for each macroinvertebrate family. In a LRM, a binary response
variable(herefamilypresenceorabsence)ismodeledasafunction
Data collection
of explanatory variables (here environmental conditions). Since
Samples were taken in the Chaguana river basin (Ecuador), multiple samples were collected on the same sampling site, the
Gilgel Gibe river basin (Ethiopia) and Cau river basin (Vietnam) responses were not mutually independent. Therefore the LRMs
(Table 1; Figure S1–S3). In each survey physical-chemical water were fitted with Generalized Estimating Equations (GEE, [28])
samples and biological samples (macroinvertebrates) were col- which account for the dependencies of the clustered sampling
lected. Per river basin, samples were taken at multiple sampling scheme. All LRMs were fitted with an independent working
sites along a pollution gradient (Table 2). Details on sampling correlation matrix.
locationsareprovidedinTable1andFigureS1–S3.Thenumber A hierarchical backward elimination model selection method
ofsamplingsitesvariedforeachriverbasin,buteachsamplingsite was carried out to build the LRM. The starting model included
wasvisitedtwiceineachyear;onceinthewetseasonandoncein five physical-chemical variables (conductivity, dissolved oxygen
thedryseason.Intotal,60,104and306samplesweretakenat15, concentration,pH,streamvelocityandwatertemperature),season
29and47samplinglocationsintheVietnamese,Ecuadorianand and river basin (represented as country). In addition to the main
Ethiopianriver basin,respectively (Table 2).Watersamples were effects, two-way interactions between the physical-chemical
analysed according to the ISO standards and only the environ- variables and the river basin as well as between season and river
mental conditions that were monitored in all three river basins basinwereincluded(TableS1andS2).Firstitwastestedwhether
were selected for further analysis, being conductivity (mS.cm21), the interactions were present at a 5% level of significance and
dissolved oxygen concentration (mg.L21), pH (-), stream velocity insignificant interactions were excluded from the model. A
(m.s21)and watertemperature (uC)(Table2). significant interaction between a physical-chemical variable and
Benthic macroinvertebrates were sampled, identified and a river basin suggests that the effect of the physical-chemical
quantified according to the method described in Gabriels et al. variable on the occurrence of the family under study differed
[6] which is an internationally accepted kick-sampling procedure betweenriverbasins.Furthermore,allmaineffectswereincluded
for macroinvertebrate sampling. A conical net with a size of independent fromtheirsignificance.
20630cm and a mesh size of 300–500 mm, attached to a stick, Residual plots and the extended Hosmer-Lemeshow test for
wasused.Withthehandnet,allaccessibleaquatichabitatswithin LRMsbasedonGEE[29]wereusedtoassessthegoodness-of-fit
a stretch of 10–20m were sampled using the kick sampling of the LRM. None of the models showed lack-of-fit. Since the
method.Thesamplingeffortwasequallydividedoverthedifferent Hosmer-Lemeshow test is insensitive to omitted quadratic terms,
habitatspersamplingsite.Theorganismswereidentifiedtofamily quadraticeffectsofthephysical-chemicalvariableswereaddedto
level and this resulted in binary presence-absence data. Subse- theLRM. However, noneof these effects were significant.
quently,fivemacroinvertebratefamilies,presentinthethreeriver The outcome of the LRM per family was visualized as the
basins and ranging from a pollution tolerant family towards a estimatedprobabilitythatthefamilywaspresentasafunctionofa
pollutionsensitivefamilybasedontheBMWPtolerancelist,were physical-chemical variable. The explanatory variables different
selected. The five target macroinvertebrate families that were from the one on the x-axis, were set to their river basin specific
selected were Chironomidae (tolerance class 2 (TS2)), Baetidae mediansandtheseasonto‘‘dry’’.Theriverbasin-specificobserved
(TS4), Hydroptilidae (TS6), Libellulidae (TS8) and Leptophlebii- range of the corresponding physical-chemical variable were
dae (TS10). For a complete overview of the taxa encountered in plotted as horizontal boxplots below the response curves and
eachriverbasinwerefertotherelatedpublications[17,18,20].No weresubdividedbetweenpresenceandabsencepoints.Thegray-
PLOSONE | www.plosone.org 2 October2014 | Volume 9 | Issue 10 | e108898
AbioticPreferencesofAquaticMacroinvertebratesintheTropics
Table1. Description ofthe studyareas.
Ecuador Ethiopia Vietnam
Riverbasinfacts
Nameofriverbasin Chaguana GilgelGibe Cau
Country SouthwestofEcuador SouthtoWestofEthiopia NorthofVietnam
Province ElOro Oromia ThaiNguyen
Tributaryriver Paguariver58 Giberiver61 Cauriver60
Source OccidentalAndes58 Ethiopianplateau62 HighlandofNorthernVietnam60
Altitude 2900m58 1096–3259m63 275m60
Surfacearea 320km258 5371km218 6030km260
Climaticdescription
Averageannualprecipitation 930mm59 2000mm63 2063mm60
Airtemperaturerange 19.9uC–31.4uC59 8.3uC–29.1uC64 10.0uC–39.0uC60
Dryseason May–November59 October–April63 October–May60
Wetseason December-April59 May–September63 June–September60
58Dominguez-Granda,2007,59Matamoros,2004,60MONRE,2006,61Demissieetal.,2013,62Uhlenbrooketal.,2010,63Broothaertsetal.,2012,18Ambeluetal.,2010,64
ColomboandMaran,2004.
doi:10.1371/journal.pone.0108898.t001
Table2. Characteristics ofbiological andphysical-chemical sampling.
Units Ecuador Ethiopia Vietnam
Numberofsamples 104 306 60
Numberofsamplinglocations 29 47 15
Monitoringyears 2005–2006 2006–2011 2009–2010
Seasons wet wet wet
dry dry dry
Physical-chemicalvariables
Streamvelocity m/s 0.560.4 0.560.3 0.560.3
Watertemperature uC 23.362.8 20.062.5 28.662.0
Conductivity mS/cm 1496152 108659 2356181
pH - 6.960.4 7.460.5 7.060.8
Dissolvedoxygen mg/L 7.161.3 6.561.5 6.460.8
Macroinvertebratesampling
Chironomidae
Numberofsamples:absent 9 49 4
Numberofsamples:present 95 257 56
Baetidae
Numberofsamples:absent 18 80 11
Numberofsamples:present 86 226 49
Hydroptilidae
Numberofsamples:absent 88 300 34
Numberofsamples:present 16 6 26
Libellulidae
Numberofsamples:absent 50 157 51
Numberofsamples:present 54 149 9
Leptophlebiidae
Numberofsamples:absent 54 300 49
Numberofsamples:present 50 6 11
Descriptivestatisticsofphysical-chemicalvariablesaregivenasmeanvalues6standarddeviations.Presence-absencerecordsperfamilyandriverbasin(representedas
country)aregivenastheamountofsamplesinwhichmacroinvertebratefamilieswerepresentorabsent,respectively.
doi:10.1371/journal.pone.0108898.t002
PLOSONE | www.plosone.org 3 October2014 | Volume 9 | Issue 10 | e108898
AbioticPreferencesofAquaticMacroinvertebratesintheTropics
Table3. P-valuesforriver basin-based interactioneffects.
Streamvelocity Watertemperature Conductivity pH Dissolvedoxygen
Leptophlebiidae 0.008 0.256 0.033 0.038 0.121
Libellulidae 0.017 ,0.001 0.767 0.055 0.015
Hydroptilidae 0.905 0.692 0.015 0.420 0.262
Baetidae 0.126 0.003 0.113 0.924 0.582
Chironomidae 0.998 ,0.001 0.999 0.999 0.999
Theeffectofanexplanatoryvariablewassignificantlydifferentbetweenthethreeriverbasinsincasethatthep-valuefortheinteractioneffectwaslowerthan0.05.
Significantrelationsareindicatedinbold.
doi:10.1371/journal.pone.0108898.t003
coloredendsoftheresponsecurvesindicateextrapolationoutside effectofconductivityandstreamvelocitywassimilarforthethree
the observed range of the corresponding physical-chemical river basins (Table S1; Figure S4–S8). Dissolved oxygen (DO)
variable. concentration and water temperature ranges differed between
Forthedataexploration,theeffectsofseasonandriverbasinon riverbasins(Table 2;FigureS9–S33).Theaveragestreamvelocity
the continuous physical-chemical variables were assessed using a washigherduringthewetseasoncomparedtothedryseason(p,
linear regression model fitted with GEE for accounting for the 0.01;TableS1;FigureS4).Theaverageconductivityinwetseason
clusteredsamplingscheme.Onlyfewp-valueswerereportedinthe was lower than in dry season in the Ecuadorian river basin (p,
main text to prevent information overload, all p-values were 0.01) and the Ethiopian river basin (p,0.01). In the Vietnamese
summarizedinTable3and4.Allstatisticaltestswereperformed river basin, however, this difference was not significant (p=0.39;
at the5%significance level. Figure S6). The seasonal effect differed per river basin for water
temperature (Figure S5), pH (Figure S7) and DO concentration
Results (FigureS8).Forinstance,dryseasonDOconcentrationstendedto
exceedwetseasonconditionsintheEcuadorianriverbasin,butin
Data exploration and correlation analysis the Ethiopian river basin the opposite was observed (Table S1;
The observed range of physical-chemical water quality condi- FigureS8).Forbothseasonstherewasnosignificantdifferencein
tions was not always equal between river basins. Seasonal effects meanstreamvelocitybetweenriverbasins(p=0.32andp=0.78,
wereobservedforallphysical-chemicalvariablesandtheseasonal Table S2; Figure S4). For DO there was a significant difference
Table4. Estimatesofthe main effects forChironomidae,Baetidae, Hydroptilidae, Libellulidae, Leptophlebiidae.
Seasonality Streamvelocity Watertemperature Conductivity pH Dissolvedoxygen
Chironomidae
Ecuador 20.873(0.265) 1.358(0.016) 0.091(0.455) 20.002(0.316) 0.149(0.697) 20.108(0.290)
Ethiopia 20.313(0.402) 1.358(0.016) 20.078(0.151) 20.002(0.316) 0.149(0.697) 20.108(0.290)
Vietnam 35.838(.0.999) 1.358(0.016) 21.649(,0.001) 20.002(0.317) 0.149(0.697) 20.108(0.290)
Baetidae
Ecuador 0.869(0.184) 1.289(0.004) 0.031(0.700) 20.002(0.121) 0.554(0.030) 20.065(0.410)
Ethiopia 0.704(0.019) 1.289(0.004) 20.049(0.407) 20.002(0.121) 0.554(0.030) 20.065(0.410)
Vietnam 0.163(0.804) 1.289(0.004) 20.351(0.133) 20.002(0.121) 0.554(0.030) 20.065(0.410)
Hydroptilidae
Ecuador 20.126(0.864) 0.829(0.223) 20.062(0.531) 20.003(0.297) 0.963(0.020) 0.179(0.519)
Ethiopia 21.463(0.203) 0.829(0.223) 20.062(0.531) 20.020(0.042) 0.963(0.020) 0.179(0.519)
Vietnam 20.408(0.438) 0.829(0.223) 20.062(0.531) 20.009(0.031) 0.963(0.020) 0.179(0.519)
Libellulidae
Ecuador 21.788(0.005) 2.488(0.005) 0.387(0.002) 20.0001(0.961) 0.674(0.006) 0.481(0.023)
Ethiopia 0.192(0.445) 20.279(0.534) 20.094(0.021) 20.0001(0.961) 0.674(0.006) 20.046(0.619)
Vietnam 21.369(0.268) 3.780(0.021) 20.708(,0.001) 20.0001(0.961) 0.674(0.006) 0.135(0.815)
Leptophlebiidae
Ecuador 2.147(0.006) 3.679(0.003) 20.291(0.048) 20.010(0.072) 0.503(0.356) 1.371(0.002)
Ethiopia 1.385(0.232) 0.694(0.750) 20.291(0.048) 20.024(0.370) 20.223(0.840) 1.371(0.002)
Vietnam 1.426(0.100) 21.703(0.410) 20.291(0.048) 0.002(0.355) 1.943(0.004) 1.371(0.002)
Correspondingp-valuesaregivenbetweenbrackets;effectsweretestedatthe5%significancelevel.Significantrelationsareindicatedinbold.
doi:10.1371/journal.pone.0108898.t004
PLOSONE | www.plosone.org 4 October2014 | Volume 9 | Issue 10 | e108898
AbioticPreferencesofAquaticMacroinvertebratesintheTropics
PLOSONE | www.plosone.org 5 October2014 | Volume 9 | Issue 10 | e108898
AbioticPreferencesofAquaticMacroinvertebratesintheTropics
Figure1.StreamvelocitydataforwhichChironomidae(A),Baetidae(B),Hydroptilidae(C),Libellulidae(D)andLeptophlebiidae(E)
arefoundtobepresent(denotedbyPontheleftaxis)andabsent(denotedbyAontheleftaxis)inEcuador(red),Ethiopia(green)
andVietnam(blue).Thesamplesizesperboxplotareshownontherightaxis.
doi:10.1371/journal.pone.0108898.g001
between river basins for the dry season (p,0.01), but not for the S39 and S42). Again, similar as for the Chironomidae, we found
wet season (p=0.16). In contrast, the mean water temperature, thatBaetidaewerelikelytobepresentinallthethreeriverbasins
pHandconductivityweresignificantlydifferentacrossriverbasins (Figure 1B,S39andS42).Intheinvestigatedrange,nosignificant
for bothseasons. association was found between the probability of occurrence of
The prevalence values, known as the relative frequencies of Baetidae and DO concentration (p=0.58, Figure S43) or
occurrence of taxa (here families, [30]), did not always approx- conductivity (p=0.11,Figure S41).
imate 0.5, i.e. some families occurred rarely or ubiquitously Hydroptilidae(TS6). AssociationsbetweentheDOconcen-
(Table 2). Furthermore, it was found that the range of preferred tration,pH,streamvelocityandwatertemperaturerelativetothe
physical-chemical conditions differed between families (Fig- probability of occurrence of Hydroptilidae were similar across
ure 1A–1E; Figure 2A–2E; Figure S9–S33). For less pollution river basins (Table3). Only the effect of conductivity differed
sensitivefamiliessuchasChironomidae(TS2)andBaetidae(TS4) between riverbasins(p=0.02;Table 4),i.e.higherconductivities
a wide range of suitable physical-chemical conditions has been were associated with a lower probability that Hydroptilidae were
observed. More sensitive families such as Leptophlebiidae (TS10) present;thiswasobservedfortheVietnameseriverbasin(p=0.03)
wereonlypresentwithinamorenarrowrangeofstreamvelocity, and the Ethiopian river basin (p=0.04; Table4; Figure S46).
conductivity and DO concentration (Figure 1A–1E; Figure 2A– Furthermore, a more alkaline pH was associated with a higher
2E;Figure S9–S33). occurrence of Hydroptilidae (p=0.02; Table4; Figure S47). No
Based on the correlation analysis no variables were discarded significant associations were found between water temperature
fromthedatasetasmostcorrelationswere‘‘weak’’sincetheywere (p=0.69; Figure S45), stream velocity (p=0.90; Figure S44) and
smallerthan0.4inabsolutevalues[21,24].Onlyconductivityand DO concentration (p=0.26; Figure S48) and the probability of
water temperature (r=0.41, p,0.01) and conductivity and DO occurrence ofHydroptilidae (Table 4).
(r=20.42,p,0.01)were moderately correlated (Table5). Libellulidae(TS8). Theassociationsbetweenwatertemper-
ature (p,0.01), stream velocity (p=0.02) and DO concentration
Logistic regression model (p=0.02)andtheprobabilityofoccurrenceofLibellulidaediffered
Chironomidae (TS2). For Chironomidae there was an between river basins (Table 3). Whereas in the Ecuadorian river
interaction effect between river basin and water temperature basin, the probability that Libellulidae occurred increased with
(p,0.01), i.e. the effect of water temperature on the occurrence increasing water temperatures (p,0.01), in the Ethiopian and
differedbetweenriverbasins(Table3).Althoughincreasingwater Vietnamese river basin inverse associations were observed
temperatures were associated with a lower probability that (p=0.02 and p,0.01, respectively; Table 4; Figure S50). Con-
Chironomidae were present in the Vietnamese river basin (p, cerningstreamvelocity,Libellulidaewerelikelytooccurathigher
0.01), in the Ecuadorian river basin and in the Ethiopian river stream velocities in the Ecuadorian and Vietnamese river basin
basinthiseffectwasnotobserved(p=0.46andp=0.15,Table4; (p,0.01 and p=0.02, respectively), but in the Ethiopian river
Figure S35). However, note that Chironomidae were absent in basin this was not observed (Table 4; Figure 4A and S49).
only4outof60samplesfromtheVietnameseriverbasin(Figure Increasing DO concentrations were associated with a higher
S35). We also found a positive association between the stream probability of occurrence of Libellulidae in the Ecuadorian river
velocity and the probability of occurrence of Chironomidae basin (p=0.02), but for the Vietnamese (p=0.82) and Ethiopian
(p=0.02),i.e.ahigherstreamvelocitywasassociatedwithahigher riverbasin(p=0.62)thisassociationwasnotsignificant(Figure4B
probabilitythatChironomidaewerepresent(Table 4;Figure 3A). andS53).TheassociationsbetweenconductivityandpHandthe
Theeffectofstreamvelocitywassimilarbetweenriverbasins,i.e. probability of occurrence of Libellulidae were similar between
regardlessthestreamvelocity,Chironomidaewerealwayslikelyto riverbasins(Table 4;Figure S51 andS52).
bepresentinthethreeriverbasins(probabilitiesbetween0.8and Leptophlebiidae (TS10). Associations between stream ve-
1.0 in Figure 3A and S34). The latter is also reflected in the locity(Figure 1E,5AandS54),pH(FigureS57)andconductivity
boxplots, which indicate that absence data for Chironomidae is (FigureS56)andtheprobabilityofoccurrenceofLeptophlebiidae
low(4outof60inVietnam,49outof306inEthiopiaand9outof were different amongst river basins (p=0.01, p=0.04 and
104 in Ecuador; Figure 1A). Within the investigated physical- p=0.03; Table3). However, when interpreting the response
chemical range, conductivity, pH, and DO concentrations were curvesitwasfoundthattheprevalenceofLeptophlebiidaewasnot
notassociatedwiththeprobabilityofoccurrenceofChironomidae balanced in the Vietnamese and Ethiopian river basin (Table2).
(Figure 3B; FigureS36–S38). ForstreamssituatedintheEcuadorianriverbasinLeptophlebiidae
Baetidae (TS4). The effect of water temperature on the weremorelikelytooccurathigherstreamvelocities(p,0.01).In
probabilityofoccurrenceofBaetidaedifferedbetweenriverbasins the Vietnamese and Ethiopian river basin, however, these effects
(p,0.01; Table3). Indeed, when looking at the response curves werenotobserved(Table 4;Figure 5AandS54),probablyduethe
(FigureS40)itisclearthatwithintheobservedwatertemperature limited amount of presence data in these river basins. The
ranges for each river basin, the slopes of the curves differ. associationbetweenconductivityandtheprobabilityofoccurrence
However,probabilitiesofoccurrencealwaysexceededthevalueof ofLeptophlebiidaesuggestedthehighestprobabilityofoccurrence
0.6 in any river basin regardless the water temperature data. atlowestconductivitiesintheVietnameseriverbasin(FigureS56).
Moreover,withinriverbasinsnosignificantassociationwasfound For all three river basins, an increase in DO concentration was
between the presence of Baetidae and water temperature positively associated with the occurrence of Leptophlebiidae (p,
(Table 4). An increased stream velocity and pH was associated 0.01; Table 4; Figure 5B and S58). Associations between Lep-
withanincreasedprobabilityofoccurrenceofBaetidaeinallthree tophlebiidae and water temperature were similar between the
river basins (p=0.01 and p=0.03, respectively; Table 4; Figure three riverbasins(Figure S55).
PLOSONE | www.plosone.org 6 October2014 | Volume 9 | Issue 10 | e108898
AbioticPreferencesofAquaticMacroinvertebratesintheTropics
PLOSONE | www.plosone.org 7 October2014 | Volume 9 | Issue 10 | e108898
AbioticPreferencesofAquaticMacroinvertebratesintheTropics
Figure 2. Dissolved oxygen concentrations for which Chironomidae (A), Baetidae (B), Hydroptilidae (C), Libellulidae (D) and
Leptophlebiidae(E)arefoundtobepresent(denotedbyPontheleftaxis)andabsent(denotedbyAontheleftaxis)inEcuador
(red),Ethiopia(green)andVietnam(blue).Thesamplesizesperboxplotareshownontherightaxis.
doi:10.1371/journal.pone.0108898.g002
Discussion analysis, Al-Shami et al. [42] also integrated geographical
variables such as longitude, latitude and altitude. Blanchette and
For pollution tolerant macroinvertebrate families (e.g. Chiron- Pearson [43] related macroinvertebrate assemblages to chloro-
omidae (TS2)) only few significant associations between physical- phyll, suspended solids, turbidity and nutrient data. Kasangaki
chemicalconditionsandtheoccurrenceofthefamilieswerefound etal.[44]studiedbenthicmacroinvertebratesinUgandaandused
(Table 4)suggestingtheiroccurrenceatawiderangeofphysical- DOconcentration,conductivity,pH,turbidityandwatertemper-
chemical conditions. This was also reflected in the data ature together with some hydromorphological stream character-
exploration as Chironomidae were present in 91%, 84% and istics.Assuch,physical-chemicalvariablesusedbyKasangakietal.
93% of the samples taken in the Ecuadorian, Ethiopian and [44] are similar to those included in our statistical analysis.
Vietnamese riverbasin, respectively (Table2).Chironomidae are Conductivitycanbeseenasageneralmeasurefordisturbanceasit
known to be tolerant to disturbance, allowing them to occur in integrates pollution related variables like minerals and inorganic
impacted environments [32]. Due to their physiological adapta- pollutants[37].Forinstance,Melo[45]concludedthatstreamsize
tions they have the ability to survive low oxygen conditions data and conductivity explained most of the variability in the
(Figure 2A and 3B, [33,34]). Because of their pollution tolerant macroinvertebrate community in a stream in the tropics. In
nature,itwassurprisingtoconcludethathighstreamvelocitiesare Colombia, Holguin-Gonzalez et al. [26] used DO concentration
associated with an increased occurrence of Chironomidae andstreamvelocitytopredictthepresenceofmacroinvertebrates.
(Figure 1A,3AandS34).However,inthisstudyweonlyidentified AccordingtoFleckerandFeifarek[46]hydrodynamics(including
macroinvertebratestofamilylevel.Chironomidaearerepresented stream velocity) play a crucial role in structuring tropic macroin-
bymanyspeciesthatshowadifferentsensitivitytoenvironmental vertebrate communities. Therefore, although only five physical-
pollution[33],whichmightexplaintheweakassociationbetween chemicalvariableswereusedinthemodels,explanatoryvariables
environmental variables and the occurrence of Chironomidae. embedded in the LRMs covered a wide spectrum of potential
Baetidae, present in 83%,74% and 82%of thesamples taken in impacts.
the Ecuadorian, Ethiopian and Vietnamese river basin, respec- Seasonal changes of environmental variables were taken into
tively (Table2), were more sensitive to pollution compared to account[10,43].Twosamplesineachyearandateachsampling
Chironomidae. The habitat preference of Baetidae was mainly location (in the wet and dry season) were collected for two
determinedbystreamvelocityandpH.Baetidaehaveoftenbeen consecutiveyearsforeachriverbasin(TableS1;FigureS4–S8).By
reported as one of the most acid-sensitive macroinvertebrate includingthese,potentialseasonaluncertaintieswereintegratedin
families [35]. As taxa sensitivity to pollution increased, more our statistical analysis.
significantassociationswerefoundbetweenenvironmentalcondi- The outcome of the LRM per family was visualized as the
tions and the occurrence of macroinvertebrate families (Table4). estimatedprobabilitythatthefamilywaspresentasafunctionofa
For instance, a significant positive association between DO physical-chemical variable (e.g. Figure 3–5). Within the physical-
concentrations and stream velocities and the occurrence of chemical ranges that were investigated and for the studied
Libellulidae (TS8) and Leptophlebiidae (TS10) was found macroinvertebrate families (see boxplots and colored zone of the
(Figure 4 and 5). DO concentration is a general indicator of response curves), nine out of twenty-five interaction effects were
water quality [34,36] and also in other studies it was found that significant(Table3).Hence,thecorrespondingvariable-macroin-
DO concentrations play a crucial role when analyzing the vertebrate relationships were different between river basins and
occurrence of macroinvertebrates (e.g., [37–39)]. In streams in suggesteddifferenthabitatpreferencesfortheinvestigatedfamilies.
Malaysia, Rawi et al. [40] found that DO concentration and For Libellulidae for instance, it was found that the effect of
stream velocity were crucial variables when explaining macroin- dissolved oxygen was different between river basins (Table3;
vertebrate diversity. For other biological communities DO levels Figure 4B).Increasing DOconcentrations were associated with a
arealsoimportant,e.g.mostfishrequireaDOconcentrationofat higher probability that Libellulidae were present in Ecuadorian
least 5 mg.L21foroptimal health [41]. river basin (p=0.02), but in the Vietnamese and Ethiopian river
Habitatsuitabilityofmacro-invertebratesprobablydependson basin this relationship was not significant (p=0.82 and p=0.62).
more factors than those included in our statistical analysis. For In the Vietnamese and Ethiopian river basin, Libellulidae were
instance, additional to the variables that were included in our lessresponsivetoshiftsinDOconcentrationsastheprobabilityof
Table5. Correlation coefficients ofthe physical-chemical variables.
Streamvelocity Watertemperature pH Conductivity Dissolvedoxygen
Streamvelocity /
Watertemperature 20.14(,0.001) /
pH 0.03(0.571) 20.21(,0.001) /
Conductivity 20.39(,0.001) 0.41(,0.001) 0.15(0.001) /
Dissolvedoxygen 0.28(,0.001) 20.28(,0.001) 0.23(,0.001) 20.42(,0.001) /
Correspondingp-valuesaregivenbetweenbrackets;effectsweretestedatthe5%significancelevel.Significantrelationsareindicatedinbold.
doi:10.1371/journal.pone.0108898.t005
PLOSONE | www.plosone.org 8 October2014 | Volume 9 | Issue 10 | e108898
AbioticPreferencesofAquaticMacroinvertebratesintheTropics
Figure 3. The probability of Chironomidae being present in relation to the stream velocity (A) and dissolved oxygen data (B)
measuredinEcuador(red,solid),Ethiopia(green,dashed)andVietnam(blue,dotdashed).Thegray-coloredendsoftheresponsecurves
indicateextrapolationoutsidetheobservedphysical-chemicalrangeinthecorrespondingriverbasin.
doi:10.1371/journal.pone.0108898.g003
occurrence hardly changed at varying DO concentrations Ecuadorian and Vietnamese river basin (p=0.01 and p=0.02,
(Table 4, Figure 4B). In the Vietnamese river basin, this can be respectively).IntheEthiopianriverbasin,thisrelationshipwasnot
explainedbytherelativefewobservedpresencesofsometaxa,i.e. observed (Table 4; Figure4A). However, a negative relationship
Libellulidaewerefoundinonlynineoutofsixtysamples(Table2). didnotmeanthatLibellulidaewereabsentintheEthiopianriver
In theEthiopian riverbasinhowever, presence andabsence data basin as there was still an estimated probability of 50% that
were more equally represented. Concerning the association Libellulidae were present within the range of observations
between stream velocity and the occurrence of Libellulidae it (Figure 4A).
was found that Libellulidae favored high currents in the
PLOSONE | www.plosone.org 9 October2014 | Volume 9 | Issue 10 | e108898
AbioticPreferencesofAquaticMacroinvertebratesintheTropics
Figure4.TheprobabilityofLibellulidaebeingpresentinrelationtostreamvelocity(A)anddissolvedoxygendata(B)measuredin
Ecuador (red, solid), Ethiopia (green, dashed) and Vietnam (blue, dotdashed). The gray-colored ends of the response curves indicate
extrapolationoutsidetheobservedphysical-chemicalrangeinthecorrespondingriverbasin.
doi:10.1371/journal.pone.0108898.g004
The fact that river basin dependent associations were found is chemical variables and macroinvertebrates present can be
not surprising. For instance, Bonada et al. [47] found that the extremely complex. Hence, this may lead to river basin specific
responseofmacroinvertebratestopollutionwasdifferentbetween associations between abiotic conditions and the biological com-
Mediterranean ecoregions. Also in the US, Zuellig and Schmidt munities.
[48] found dissimilar regional benthic invertebrate community Since the range of the observed physical-chemical conditions
compositions. Moreover, it was stated by Thorne and Williams were not always equal between river basins, in some cases
[13] that due to untreated domestic and urban effluents in extrapolationsoutsidetheobservedrangewereshown(Figure 4A).
developingcountriestherelationshipbetweenindividualphysical- For instance, water temperature measurements significantly
PLOSONE | www.plosone.org 10 October2014 | Volume 9 | Issue 10 | e108898