The cumulative impacts of small reservoirs on hydrology: A review

2018·2.7 MB·English

by  Florence Habets

ScienceoftheTotalEnvironment643(2018)850–867
ContentslistsavailableatScienceDirect
Science of the Total Environment
journal homepage: www.elsevier.com/locate/scitotenv
Review
The cumulative impacts of small reservoirs on hydrology: A review
FlorenceHabetsa,*,JérômeMolénatb,*,NadiaCarluerc,OlivierDouezd,DelphineLeenhardte
aCNRS/SorbonneUniversité,UMR7619Metis,Paris,France
bUMRLISAH,UnivMontpellier,INRA,IRD,MontpellierSupAgro,Montpellier,France
cIrstea,URRiverLy,centredeLyon-Villeurbanne,Villeurbanne69625,France
dBRGM,Bordeaux,France
eToulouseUniv,INRA,INPT,INPEIPURPAN,AGIR,CastanetTolosan,France
H I G H L I G H T S G R A P H I C A L A B S T R A C T
• The number of small dams is still
increasing and is approaching
39damspersquarekilometre.
• Small dams lead to a decrease in
annualstreamdischargeof13%±8%.
• Cumulative impacts cannot be esti-
matedusingsimpleindicators.
• Cumulative impacts are difficult to
estimateandaremostoftenquanti-
fiedfrommodelling.
• The lack of information on small
reservoir characteristics is a real
shortcoming for properly estimating
theircumulativeimpacts.
A R T I C L E I N F O A B S T R A C T
Articlehistory: Thenumberofsmallreservoirshasincreasedduetotheirreducedcost,theavailabilityofmanyfavourable
Received1February2018 locations,andtheireasyaccessduetoproximity.Thecumulativeimpactsofsuchsmallreservoirsarenot
Receivedinrevisedform12June2018 easytoestimate,evenwhensolelyconsideringhydrology,whichispartiallyduetothedifficultyincollecting
Accepted15June2018 dataonthefunctioningofsuchreservoirs.However,thereisevidenceindicatingthatthecumulativeimpacts
Availableonline4July2018
ofsuchreservoirsaresignificant.
Theaimofthisarticleistopresentareviewofthestudiesthataddressthecumulativeimpactsofsmall
Editor:AshanthaGoonetilleke
reservoirsonhydrology,focusingonthemethodologyandonthewayinwhichtheseimpactsareassessed.
Mostofthestudiesaddressingthehydrologicalcumulativeimpactsfocusedontheannualstreamdischarge,
Keywords: withdecreasesrangingfrom0.2%to36%withameanvalueof13.4%±8%overapproximately30references.
Smallreservoirs
However,itisshownthatsimilardensitiesofsmallreservoirscanleadtodifferentimpactsonstreamdis-
Cumulativeimpacts
Hydrology chargeindifferentregions.Thisresultisprobablyduetothehydro-climaticconditionsandmakesdefining
Waterresourcemanagement simpleindicatorstoprovideafirstguessofthecumulativeimpactsdifficult.Theimpactsalsovaryintime,
Anthropization withamoreintensereductionintheriverdischargeduringthedryyearsthanduringthewetyears.This
Review findingiscertainlyanimportantpointtotakeintoconsiderationinthecontextofclimatechange.
Twomethodsaremostlyusedtoestimatecumulativeimpacts:i)exclusivelydata-basedmethodsandii)
models.Theassumptions,interestsandshortcomingsofthesemethodsarepresented.Scientifictracksare
proposedtoaddressthefourmainshortcomings,namelytheestimationoftheassociateduncertainties,
thelackofknowledgeonreservoircharacteristicsandwaterabstractionandtheaccuracyoftheimpact
indicators.
©2018ElsevierB.V.Allrightsreserved.
*Correspondingauthors.
E-mail addresses:fl[email protected](F.Habets),[email protected](J.Molénat).
https://doi.org/10.1016/j.scitotenv.2018.06.188
0048-9697/©2018ElsevierB.V.Allrightsreserved.
H.Florenceetal. /ScienceoftheTotalEnvironment643(2018)850–867 851
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 851
2. Evidenceoftheimpactsofsmallreservoirsonhydrology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 852
3. Howdosmallreservoirsimpacthydrology? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853
3.1. Waterbalanceofasmallreservoir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853
3.2. Lossesfromsmallreservoirs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853
3.2.1. Seepage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853
3.2.2. Evaporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855
3.3. Connectiontothestream . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855
4. Methodstoestimatethecumulativeimpactsofsmallreservoirsonhydrology . . . . . . . . . . . . . . . . . . . . . . . . . . 855
4.1. Exclusivelydata-basedmethods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855
4.1.1. Fromobservationofselectedreservoirstoestimationofcumulativeimpacts . . . . . . . . . . . . . . . . . . . . 855
4.1.2. Statisticalanalysesoftheobserveddischarge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855
4.1.3. Paired-catchmentexperiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856
4.2. Modellingapproaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857
4.2.1. Reservoirwaterbalancemodel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857
4.2.2. Reservoirinflowquantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857
4.2.3. Reservoirspatialrepresentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857
4.2.4. Aggregaterepresentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857
4.2.5. Statisticalrepresentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 858
4.2.6. Distributedrepresentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 858
5. Howtoobtainaccesstotheinformationneededonsmallreservoirs? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 859
5.1. Whattypeofdata? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 859
5.2. Physicalandtopographicalcharacteristicsofsmallreservoirs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 860
5.3. Waterreservoirmanagementcharacteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 860
6. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 861
6.1. Theuncertaintyissue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 861
6.2. Improvingknowledgeofsmallreservoircharacteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 861
6.3. Improvingabstractionestimations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 862
6.4. Impactindicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 862
7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 862
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 863
AppendixA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 863
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864
1. Introduction
(Berg et al., 2016; Mbaka and Wanjiru Mwaniki, 2015; Downing,
Large reservoirs have strong impacts on hydrology at regional 2010;PoffandZimmerman,2010;St.Louisetal.,2000),ashavethe
toglobalscales.Indeed,itwasestimatedthatsuchlargereservoirs impacts of such reservoirs on hydrology (Nathan and Lowe, 2012;
have led to a approximately 2% (Biemans et al., 2011), to a sea Fowleretal.,2015).Thereportedimpactsaregenerallystrongbut
leveldecreaseofapproximately30mm(Chaoetal.,2008),andthat presentalargevariation.
theystoreavolumeequivalenttoapproximately10%ofthenatural Estimatingthecumulativeimpactsofsystemsofsmallreservoirs
annual soil storage capacity at the global scale (Zhou et al., 2016). onagivenbasinhasbecomeanissueastheirnumberincreases(for
However,thesestudiesdidnotconsidertheimpactsofsmallerreser- instance, a 3% increase per year in the US; Berg et al., 2016). This
voirs on hydrology. Downing (2010) found that small ponds and trendmaypersistbecausethesesystemsareoftenconsideredtobea
lakes(smallerthan0.1km2)coveralargerareaandaremorenumer- techniquetoadapttoclimatechange(vanderZaagandGupta,2008).
ous than large reservoirs and that approximately 10% of them are Indeed,smallreservoirsaremainlyusedtostorewaterduringthe
constructedreservoirs. wetseasontosupportwateruseduringthedryseason,particularly
When considered individually, each reservoir may modify its forirrigationandlivestockinruralareas(Wisseretal.,2010;Nathan
local and remote environment. The cumulative impacts of many andLowe,2012);tostorewaterduringstormstopreventflooding;
reservoirs in a catchment are the modifications induced by a set ortostoresedimentsincheckdamstoreduceerosionandmuddy
ofreservoirs(orreservoirnetwork)takenasawhole.Thecumula- floodrisks.Becausethepartoftheglobalpopulationthatwillexperi-
tiveimpactsarenotnecessarilythesumofindividualmodifications encewaterscarcityisprojectedtoincreasewithclimatechangeand
becausereservoirsmaybeinter-dependent,suchascascadingreser- becausetheintensityofstormeventsisalsoprojectedtosimultane-
voirsalongastreamcourse.Cumulativeimpactsarenotthesimple ouslyincrease(Pachaurietal.,2014),thereisincreasingpressureto
additionofindividualimpacts:theycandevelopviaanadditiveor constructsmallreservoirs(vanderZaagandGupta,2008;Thomaset
incremental process, a supra-additive process (where the cumula- al.,2011).
tive effect is greater than the sum of the individual effects) or an However,anuncontrolleddevelopmentofsuchsmallreservoirs
infra-additiveprocess(wherethecumulativeeffectislessthanthe mayincreasethewaterresourceprobleminbothquantitativeand
sum of the individual effects). The total impact is therefore equal qualitativeways.Thus,watermanagersareseekingsomeindicators
to the sum of the impacts of the developments and to interaction that would help to determine optimal networks of small reser-
effects. Indeed, addressing the cumulative impacts implies cover- voirs in terms of storage capacities and in terms of locations and
ingdifferentspatialandtemporalscales(CanterandKamath,1995) management. Consequently, in France, the Ministry of the Envi-
and having a reference state (McCold and Saulsbury, 1996). The ronment requested a joint scientific assessment to collect useful
cumulativeimpactsofsmallreservoirsonsedimenttransport,bio- information/knowledgeandtoolstoprovidelocalstakeholderswith
chemistry,ecologyandgreenhousegasemissionshavebeenstudied such indicators and methods to assess the cumulative impacts of
852 H.Florenceetal. /ScienceoftheTotalEnvironment643(2018)850–867
smallreservoirs.Thisrequestledtoareviewcoveringbiochemistry, TherightpartofFig.1showsthattheimpactsonannualflows
ecology, hydrology and hydromorphology (Carluer et al., 2016). In arenotconstantfromyeartoyearbuttendtobelowerduringthe
this paper, a full review of the cumulative impacts of small reser- wet years and two times greater than the median impact in the
voirsonhydrologyispresentedbecausethehydrologicalimpactwill driestyears.Thisresultisveryimportantbecauseitindicatesthat
affecttheotherimpacts.Althoughthereisnoaccepteddefinitionof evenwithoutchangingthesmallreservoirnetwork,itsimpactswill
small reservoirs, it is commonly accepted that the storage capaci- changeinthecontextofclimatechange:itmaydecreaseinareasthat
tiesofsuchreservoirsarebelow1millionm3,asstatedbyAyalewet willbecomewetterbutmayincreaseinareasthatwillbecomedrier.
al.(2017)andThomasetal.(2011).Thisreviewdoesnotextendto One key issue in estimating the cumulative impacts is under-
theverysmallreservoirsoffewhundredsofm3thatcanbeusedfor standinghowsuchimpactsarerelatedtothereservoirnetwork,i.e.,
waterharvesting(LasageandVerburg,2015). thelevelatwhichthebasinisequippedwithsmalldamstoavoid
First,asynthesisofthequantificationoftheimpactsatthebasin over-equippingthebasin,withconsequencesintermsofeconomy
scale is presented, and the ability of some conventional descrip- and ecology. Having a single indicator or a set of indicators capa-
tors to be used as indicators is studied. Then, the various ways in bleofestimatingthecumulativeimpactsofsmallreservoirsonthe
which small reservoirs can impact the water cycle are presented, meanannualdischargewouldbehelpfultomostwatermanagement
alongwiththemethodsthatareusedintheliteraturetoestimate agencies.Basedontheestimatedvaluescollectedintheliterature,
the cumulative impacts of such numerous and not always well- apreliminaryanalysiswasperformedtodeterminewhethersome
knownstructures.Theseresultsarethendiscussed,addressingthe easy-to-accesspropertiesofthereservoirnetworkcouldbeusedas
uncertainties,long-termtrends,andimpactsonotherbiochemical, indicators.Forthispurpose,wecollectedthemaincharacteristicsof
ecologicalandsocialcomponents. the basins and of their small reservoir network from the available
studiesandattemptedtoconnectthemtotheimpactsonthemean
2. Evidenceoftheimpactsofsmallreservoirsonhydrology annualdischarge.Weusedthereservoir’sdensity,expressedasthe
numberofreservoirspersquarekilometreorasthevolumestored
Fromtheliteraturereview,thecumulativeimpactsofsmallreser- persquarekilometre,andthemeanprecipitationorthemeandis-
voirs on hydrology are most often estimated from the annual dis- chargeinthebasin.TheresultspresentedinFig.2showthatnoneof
charge,lowflowsandfloods.Thereisageneralconsensusthatsetsof thesecharacteristicsareabletobeusedasindicatorsforsuchcon-
smallreservoirsleadtoareductioninthefloodpeaks(Frickel,1972; trastedbasinsastheonesfoundintheliterature.Indeed,withina
Galeaetal.,2005;NathanandLowe,2012;Thompson,2012;Ayalew narrowrangeofspecificdischargeorprecipitation,thedecreaseof
etal.,2017)ofupto45%,particularlysincesomereservoirsarecon- the annual discharge varies a lot and can not be correlated to the
structedasstormwaterretentionponds(Fennesseyetal.,2001;Del densityofreservoirnetwork.
Giudiceetal.,2014).However,over-toppingfloodingordamfailure Amoreregional-scaleviewcouldbeusefultoattempttodisen-
can result in large floods (Ayalew et al., 2017), which may lead to tangledifferenttypesofclimateoruse.However,accordingtothe
casualtiesincludingdeath(Tingey-Holyoak,2014).Suchfailurescan sample of available studies, only a continental-scale analysis was
bemorefrequentforsmalldamsthanforlargerdamsduetothelack possible.Itappearsfromthesefiguresthatthegeneralcharacteris-
of adapted policies, which may lead to a lack of maintenance and ticspresentawiderspreadbetweencontinentsthanwithinagiven
a tendency to store excess water to secure production (Pisaniello, continent,eveniftheresultsarefromdifferentstudies.Forinstance,
2010;CamnasioandBecciu,2011;Tingey-Holyoak,2014). thespecificdischargeislowinAustralia,thedensityislowinAfrica,
Thelowflowsarealsofrequentlyreportedtodecreasewhenaset andthestoragevolumetendstobeimportantinAmerica.However,
ofsmallreservoirsispresentinabasin(Nealetal.,2000;O’Connor, even within a continent, these characteristics are not sufficiently
2001;HughesandMantel,2010;NathanandLowe,2012;Thompson, welllinkedtotheimpactstobereliablyusedasindicators.
2012) with a large spread (0.3 to 60%), although the water stored This result occurs because the cumulative impacts of reservoir
canoccasionallybeusedtosustainalowflow(Thomasetal.,2011). networks rely on a large number of factors: the hydrological pro-
Themajorityofstudieshavefocusedontheannualstreamdischarge, cesses occurring in each reservoir, the water management (water
reportingadecreaseinthemeanannualdischargethatrangesfrom abstractionrateandtiming,wateruptakesfromandreleasestothe
0.2%(HughesandMantel,2010)to36%(Meigh,1995).Onaverage, river), the reservoir characteristics, the reservoir network geome-
in approximately 30 references, the decrease in the mean annual try, and the connectivity of each reservoir to the stream drainage
dischargereaches13.4%±8%(Fig.1andAppendixTableA.1). network.Thesepointsaredetailedbelow.
Fig.1. Left:Distributionoftheestimatedannualstreamdischargedecreaseattributedtoreservoirnetworks.Thedistributionisestablishedbasedon20values.Right:Impact
ontheannualdischargeestimatedduringwet,mediananddryyears.Eachbarcorrespondstoadifferentcatchment.Theestimationsarefromthefollowingreferences:a:
Gutteridge-Haskins-Davey(1987),b:OckendenandKotwicki(1982),c:DubreuilandGirard(1973),d:Cresswell(1991),e:Teoh(2003),f: Habetsetal.(2014),andg:Kennon
(1966).
H.Florenceetal. /ScienceoftheTotalEnvironment643(2018)850–867 853
Fig.2. Cumulativeimpactsofthesmallreservoirsonthemeanannualdischarge(colourscaleontheright),estimatedfromstudiesreportedinAppendixTableA.1,asafunction
ofpossibleindicators:reservoirdensityexpressedasthenumberofdamspersquarekilometreandasstoragecapacityincubicmeterpersquarekilometre,annualprecipitation
expressedinmm/m2/year,orspecificdischargeexpressedinmm/m2/year.Eachpointrepresentsacatchment,andthesymbolcorrespondstodifferentregions:Africa,America,
Asia,andAustralia.
3. Howdosmallreservoirsimpacthydrology? agroundwaterinflow,althoughnonewasreportedintheliterature
review.
Smallreservoirshaveanimpactonhydrologybecausetheyaffect Outfluxincludesoutflow(downstreamflow)andwaterabstrac-
the natural water cycle that would occur without reservoirs. To tion,aswellasevaporationandseepagelossesfromthereservoir.
understand how networks of small reservoirs impact river flow at Outflowisdefinedasthedownstreamflowduetoreservoirrelease.
thebasinscale,itisnecessarytounderstandthefunctioningofasin- Abstractioncorrespondstothewateruptake,oftenbypumping,for
glereservoir,howitcanhaveanimpactontheriverflowandwhy humanuse(irrigation,livestockwatering,andsoforth).Seepageflow
theimpactvariesintimeandfromonereservoirtoanother. mayoccuraswaterinfiltrationthroughthereservoirbedorthrough
orbelowthedam.
All these fluxes can vary considerably from one reservoir to
3.1. Waterbalanceofasmallreservoir
another.Forinstance,abstractioncanbethemainoutput,especially
for farm reservoirs. However, it can also be null, such as in storm
Fig.3presentsthevarioustermsofthewaterbalanceofthereser-
waterorcheckdamreservoirs.Section6.3discusseshowabstraction
voir.Fromageneralperspective,thereservoirwaterbalancecanbe
canbeestimatedatthebasinscale.
expressedbythefollowingequation:
Waterlossesarepresentforeverytypeofreservoir,butwitha
largespreadofintensity,rangingfromthemainoutfluxtonegligible
ddVt =Qin+P+GWin−Qout−E−S−Qabs. (1) obneeess.tTimheatneedx.tsectionfocusesontheselossesandonhowtheycan
Here, dV is the water volume variation [m3] over the period 3.2. Lossesfromsmallreservoirs
dt[s],Qinisthestreaminflowtothereservoir[m3/s],Qoutistheout-
flowfromthereservoir[m3/s],Eistheevaporationrate[m3/s],Pis 3.2.1. Seepage
theprecipitationrate[m3/s],Sistheseepagerate[m3/s],GW isthe Seepage(alsocalledpercolationflux)maybeparticularlyimpor-
in
groundwaterinflow[m3/s]andQ isthewaterabstraction[m3/s]. tanttoconsiderforsmallreservoirsbecausemostofthesereservoirs
abs
Inflowcanhave4sources:i)theupstreamflow,whichdepends are built with earthen dams. The seepage rate depends on the
on the way in which the reservoir is connected to the river hydraulic head gradient betweenthe reservoir and the underlying
(Section3.3);ii)thesurfacerunofffromtheareadirectlydrainedby aquifer(orunsaturatedzone)ordamwall,aswellasonthehydraulic
thereservoiralongitsbank;iii)theinterceptedprecipitation;andiv) conductivitiesoftheaquiferandreservoirbedmaterial.
854 H.Florenceetal. /ScienceoftheTotalEnvironment643(2018)850–867
Fig.3. Waterbalanceofasmallreservoiranditsmaindrivers.Thecomponentsofthewaterbalanceareindicatedbylargearrows:inputscanbeinflows,suchasupstreamrunoff,
lateralsurfacerunoff,anddirectprecipitation;outputscanbeoutflows,abstraction,seepageandevaporation.
Althoughseepageisalossatthereservoirscale,thewaterisnot Whenthecumulativeimpactsareconsidered,boththeseepage
lost and is mostly diverted. Indeed, infiltration tanks, encountered rate and the seepage fate are important. In the case of infiltra-
especiallyinAsia,arebuilttofavourinfiltrationthroughthereservoir tionintothedamwall,theseepagewatermightflowdownstream
bedtoincreasethegroundwaterrecharge.Inthisway,alargerpart in the river, and thus, the seepage flux might not be lost at the
ofthemonsoonflowisstoredinthegroundwaterwhileavoidingthe scale of the river basin. An illustration of such a process was pro-
evaporationlossfromreservoirsduringthedryseason(Glendenning videdbyKennon(1966),whoobservedthatephemeralrivershave
etal.,2012).However,whendamsareintendedtostorewaterover becomepermanentaftertheimplementationofdamsbuilttopre-
thelongterm,seepageisconsideredasaloss.Insuchcases,imper- vent erosion (see Section 4.1.1), and by those studies that include
viouslayersofclayorgeomembrane(Alonsoetal.,1990;Yiasoumi groundwater recharge from dam seepage (Ramireddygari et al.,
and Wales, 2004) are used to reduce seepage, but their efficiency 2000; Barber et al., 2009; Smout et al., 2010; Shinde et al., 2010;
decreaseswithage.Thus,irrespectiveoftheintendedfunctionofthe Perrin et al., 2012). Therefore, seepage fluxes from each reservoir
reservoir,itisratherimportanttoestimatetheseepageratefromthe shouldnotbeaggregatedtoestimatethelossatthebasinscaleand
reservoirbecauseitdeterminesitsefficiencyforstoringwater(then, thusfortheestimationofthecumulativeimpactsofsmalldamson
alowseepagerateisexpected)orwithinthegroundwater(then,a hydrology.
highseepagerateisexpected).Intheliterature,estimationsofthe
seepageratewerebasedonwaterbalanceapproachesconstrained
bylocalobservationsoftheprecipitation,potentialevaporationand
reservoir’s water level (Culler, 1961; Kennon, 1966; Sukhija et al.,
1997; Singh et al., 2004; Bouteffeha et al., 2015), as well as on
additionalobservationsofthesoilmoistureandpiezometricheads
(Shinogi et al., 1998; Antonino et al., 2005; Massuel et al., 2014),
environmentaltracers(Sukhijaetal.,1997),ormorefrequentlyon
modelling approaches (Zammouri and Feki, 2005; Boisson et al.,
2014;JainandRoy,2017).
Fig.4presentssomeestimationsoftheseepageandevaporation
lossesfromtheliteratureunderdifferenthydroclimaticcontextsand
forreservoirsbuiltforvariouspurposes.Mostestimatedseepageval-
ues are greater than 5mm/day on average in the studied periods,
andthus,theseepagerateappearstobehigherthantheevapora-
tion rate. However, most of the values found in the literature are
frompercolationtanks,i.e.fromdamsbuilttopromotearapidinfil-
tration of the runoff during the wet season to recharge the water
table.Fortheothertypesofdams,theestimationscanbelower:less
than 1mm/day for Culler (1961) in the US and up to 6.2mm/day
forShinogietal.(1998)overa6-monthperiodinabasininBrazil.
Fig.4. Estimationoftheseepagelossandtheevaporationfluxofsmallreservoirs
Fowleretal.(2012,2015)considerthathillslopedamsinAustralia
onaseasonaltoannualbasis.Twotypesofreservoirsaredistinguished:infiltration
arenotefficientforstoringwateriftheseepagerateisgreaterthan
reservoirsandothertypesofreservoirs.Thevaluesaretakenfromthearticlescitedin
5mm/day. thissection.
H.Florenceetal. /ScienceoftheTotalEnvironment643(2018)850–867 855
3.2.2. Evaporation flowwhileupstreamflowexists,suchasforhillslopereservoirsand
Unlikeseepage,evaporationfluxesfromeachreservoirshouldbe check dams. Such reservoirs have strong impacts on the intensity
aggregated at the basin scale. The impact of the reservoirs on the andthedurationoflowflows.Inparticular,theresumptionofflow
evaporation losses is then the difference between the evaporation inthefallcanbesignificantlydelayed.Inthecaseofdiversionora
fromthelandcoverthatwaspresentpriortothedamsbeingbuilt minimumflowbypassreservoir,adownstreamflowisensuredwhen
and the evaporation from the reservoirs. Such estimations are not the upstream flow is non-zero. If the reservoir is located in diver-
straightforward,particularlybecausetheheatstorageofthewater sion,thenthefillingperiodofthereservoircanbemanagedsuchthat
bodyaffectsthesurfaceenergyflux(Assoulineetal.,2008;McMahon thereservoirmayhavenoimpactontheriverflowduringpartsof
et al., 2013). This storage partly depends on the temperature of theyear,whichmayallowpreservingtheecologicalfunctionofthe
the water columns, which is impacted by the depth of the dams river.Thismanagementcanalsobeadaptedtothehydrologicalsitu-
(although in opposite ways depending on the references (Girard, ationofeachyear.Thereservoirsbuiltmainlytofavourgroundwater
1966;MartínezAlvarezetal.,2007;Maglianoetal.,2015)duetothe rechargecanhavealltypesofconnectionswiththeriver;however,it
associatedchangeinthefreewaterarea);onthewatercirculation appearsthatmostofthemarebuiltdirectlyintheriverstream,thus
withinthereservoir(whichisalsoimpactedbythereservoir’sman- collectingalltheupstreamflows(Shinogietal.,1998;Sideriusetal.,
agement);andontheinteractionwiththeedges,whichcanberather 2015).Dependingontherespectiveinflowandabstractiondynamics,
closeforsmallreservoirsandthataffectsthewindvelocityandthe cumulativeabstractionmayexceedthereservoirstoragecapacity,as
advectionofairhumidity(Fig.3).Severalmethodswereusedtopro- illustratedinFig.5forexample,forwhichtheabstractionsfromthe
videestimationsoftheevaporationfromsmallreservoirsbasedon reservoirsreach105to120%ofthemaximumstoragecapacity.
observations: energy balance approaches (Anderson, 1954; Culler,
1961; Kennon, 1966; Gallego-Elvira et al., 2010), eddy-covariance 4. Methodstoestimatethecumulativeimpactsofsmall
measurements(Rosenberryetal.,2007;Tannyetal.,2008;Mengistu reservoirsonhydrology
andSavage,2010;Nordboetal.,2011;McJannetetal.,2013),scin-
tillometers(McJannetetal.,2013;McGloinetal.,2014),andwater Quantifyingthecumulativeimpactsofsmallreservoirshasbeen
balance approaches (Girard, 1966; Martínez Alvarez et al., 2007). conducted using a variety of methods, all of them requiring data
Fig. 4 presents the estimations found in the literature. The mean and observations. Two classes of methods can be distinguished: i)
annualestimationsrangefrom1.4to5.5mm/day,andthereported themethodsexclusivelybasedontheanalysisofobserveddataand
summervaluesareallabove3mm/day. ii)themethodsbasedonhydrologicalmodelling.
MartínezAlvarezetal.(2007)proposedarelationshipbetween
thesmallreservoirevaporationlossandtheClassApanevaporation 4.1. Exclusivelydata-basedmethods
thatvariesaccordingtothereservoir’sdepthandareaandthatvaries
intime(from86to94%). 4.1.1. Fromobservationofselectedreservoirstoestimationof
Severalestimationsofthesmallreservoirevaporationlossbased cumulativeimpacts
on meteorological data were proposed (de Bruin, 1978; Martínez Thisapproachwasmainlyusedinearlyworksperformedfrom
Alvarez et al., 2007; McJannet et al., 2013; McMahon et al., 2013; the50stotheearly70sintheUS(Kennon,1966;Culler,1961;Frickel,
Morton,1983).BenzaghtaandMohamad(2009),MartínezAlvarezet 1972)andinBrazil(Dubreuiletal.,1968;DubreuilandGirard,1973;
al.(2008)andCraig(2008)foundthattheevaporationlossesfrom Molle,1991).Inlightofthesepioneeringworks,itcanbeobserved
reservoirscanbeveryimportantattheregionalscaleandhavean thatthecumulativeimpactsonhydrologyhavebeenascientificand
importanteconomicimpact. watermanagementissueforalongtime.
Several techniques might help reduce evaporation from reser- Despitesomedifferencesinthemethodologyamongthesestud-
voirs: casual chemical treatment to modify the albedo or form a ies, they all aimed at quantifying single reservoir hydrologic func-
monolayerfilm,completelyorpartiallycoveringthereservoirs,man- tioningfromthemonitoringofasampleofreservoirs.Losseswere
agingthereservoiredgestoreducewindspeed,andoptimizingthe estimatedusingamassbalanceofthesampledrepresentativereser-
useofthewaterinreservoirnetworksbasedonthetemperatureof voirs based at least on the monitoring of the water level, inflows
thewaterinthereservoirs(Barnes,2008;Lund,2006;Assoulineet andoutflowsofthereservoirs.Theseearlystudiesinitiallymadethe
al.,2011;Martínez-AlvarezandMaestre-Valero,2015;Gallego-Elvira assumptionthatcumulativereservoirimpactswerethesumofthe
etal.,2011;Carvajaletal.,2014;Recaetal.,2015).However,such impactofeachreservoirfollowinganaggregationprocess.However,
techniques are not yet widely used and are not considered in the themainoutcomeofthesestudieswastoshowthatthisassumption
existingcumulativeimpactstudies. wasnotvalid.Indeed,Culler(1961)andKennon(1966)foundthat
the seepage was a significant loss for the sampled reservoirs but
3.3. Connectiontothestream contributed to downstream flow. Therefore, interactions between
reservoirsandhydrologiccompartments,especiallythestream,were
Byitself,theconnectionofthereservoirtothestreamiskeyto identifiedveryearlyasprocessestobetakenintoconsiderationto
understandingtheimpactsofthereservoirontheriverflow.Indeed, reliablyestimatethecumulativeimpacts.
thisconnectionwillimpactboththeinflowandtheoutflow.Small
reservoirs can collect all the upstream flow (Fig. 5a for a hillslope 4.1.2. Statisticalanalysesoftheobserveddischarge
reservoir or dam situated on the stream with no minimum flow) The idea is to connect the detected changes in the statistical
or only a part of the flow (reservoir with minimum flow by-pass, properties of river discharge time series with the evolution of the
Fig.5b,whichallowsmaintainingaminimumflow,ordamsituated reservoirnetworkwithinthebasin.Indoingso,thedetailsofeach
indiversionFig.5csinceinthiscase,thereservoircannotfillaslong reservoirfunctioningarenottakenintoconsideration.Toourknowl-
asinflowdoesnotexceedsomethresholds).Inthecasethatallthe edge, this type of study based solely on observations was only
upstreamflowsarecollected,thedownstreamoutflowwillprimar- performedbyGaleaetal.(2005).Astudybasedona30-yearriverdis-
ilydependonthelevelofthespillandonthereservoirwaterstorage. chargetimeseriesoftwoFrenchcatchmentsshowednostationarity
Following“fill-and-spill”(Deitchetal.,2013),downstreamdischarge breakinsummer,whileabreakwasshowninwinter,i.e.,duringthe
occursonlywhenthereservoirisfullyfilled;conversely,aslongas fillingperiod(Galeaetal.,2005).
thereservoirhasnotreacheditscapacity,downstreamdischargeis One difficulty of such statistical analyses is discriminating the
null. Therefore,it is possibleto have periodswithnodownstream specificimpactofsmallreservoirsfromthoseoflanduseandland
856 H.Florenceetal. /ScienceoftheTotalEnvironment643(2018)850–867
Fig.5. Illustrationof3differentconnectionsbetweentheriverandthereservoiranditsconsequencesintermsofriverflow.Inflow,outflowandabstractionareaccumulated
weeklyvalues,whereasstorageisaweeklyvalue.Theyareallexpressedasafractionofthemaximumstorage.Abstractionsinthereservoirsreached105to120%ofthemaximum
storagecapacity.a)Hillslopereservoirismanagedasafillandspill,withaweakandirregularinflow.b)Theminimumflowbypassensuresthataminimumoutflowoccursas
longasinflowispresent.c)Thereservoirindiversionisexpectedtofillupassoonastheinflowreachesagivenminimumflowordependingonmanagementpractices.
cover (LULC) evolution or of climate change (CC). Reservoir devel- 4.1.3. Paired-catchmentexperiment
opment occurred over decades, a sufficiently long period to be A paired-catchment experiment is an approach already used
sensitivetoLULCmodifications(suchasagriculturalintensification in hydrology for quantifying the impact of LULC changes from a
orcrop modification)and CC. Toovercomethisissue,Schreideret comparativeanalysisofstreamflowsmonitoredintwocontrasted
al. (2002) compared the observed river flows with simulated ones catchments (see, for instance, Brown et al., 2005 for a review in
obtainedusing the observedatmosphericforcing,but withoutany foresthydrology).Thompson(2012)is,toourknowledge,theonly
explicitrepresentationofthesmalldamsinthemodels.TheIHACRES studyusingthisapproachtocomparestreamflowsfromtwoadja-
rainfall-runoff model, a dynamic, lumped parameter model, was centandsimilarcatchments,onewithoutareservoirandthesecond
usedtosimulatestreamflowwithparameterscalibratedconsidering withthreesmallreservoirs.Froman18-monthmonitoring,annual
periods before the development of reservoirs. They found signifi- streamflowwasestimatedtobelowerby40%inthecatchmentwith
cantdecreasingtrendsintheobserveddischargeofbasinsthathad 3reservoirsthaninthe“no-reservoir”catchment(Thompson,2012).
adevelopmentoffarmdamcapacity,andtheywereabletoattribute Althoughtheexperimentfounddifferencesinthespecificdischarge,
these trends to non-climatic stressors since such trends were not thefullcomparisonofthewaterbalanceremaineddifficult.Themain
simulatedwithareservoir-freebasin. shortcomingofThompson’sapproachisthatcatchmentproperties
H.Florenceetal. /ScienceoftheTotalEnvironment643(2018)850–867 857
(soils,lithology,landcover,topography,andsoforth)werespatially 4.2.2. Reservoirinflowquantification
heterogeneousoverashortdistance,makingdecipheringthestream Inmostmodellingapproaches,upstreaminflowisprovidedbya
flowdifferencesdifficult.Furthermore,indirectreservoirimpactson catchmenthydrologicalmodelsimulatingthewaterbalance(WB),or
landuse,suchasthecattlegrazingaroundthereservoirinThomp- theenergyandwaterbalance(EWB),intheupstreamcatchmentand
son’s case study, can also modify stream flow. The study would the routing of the flow downstream (Table 1). Existing catchment
havebenefitedfromfollowingtheclassicapproachusedinpaired- hydrologicalmodelsareusedinthemodellings,reflectingthediver-
catchment experiments, implying a calibration period where both sityofcurrenthydrologicalmodels.Suchmodelsneedatmospheric
catchments are monitored, followed by a period when one of the forcingandsomeinformationonthelandcover,soilandtopogra-
catchmentsissubjectedtolandusechange(reservoirbuilding)and phy, unless the model parameters are calibrated without any data
the other remains as a control. However, building a reservoir net- onthephysiographiccharacteristics.Twomodels,TEDIandDeitch
workoveralargeareaisgenerallydifficultforpracticalandfinancial (Table 1), developed an alternative and pragmatic method based
reasons.Consequently,suchanapproachhasneverbeenutilizedto on using observed discharge time series as input to the model. In
ourknowledge. doingso,theTEDIandDeitchmodelsdonotbelongtoanycurrent
modelling approaches. Using observed discharge at available river
gaugesimpliesbeingabletosuccessivelyi)disentanglethenatural
4.2. Modellingapproaches flowfromtheanthropogenicflowandii)distributetheobserveddis-
chargealongthereservoirnetworks.Toachievethefirststep,Deitch
Modelling is the most widely used approach for studying and etal.(2013)usedhistoricalgaugeddischargemeasuredpriortothe
quantifying the cumulative impacts of small reservoirs. Although reservoirpre-developmentperiod.Thedischargewasthenspatially
various modelling approaches have been developed, all are based distributedaccordingtothedrainageareaofeachreservoirandthe
onthecouplingofthesmallreservoirwaterbalancemodelwitha spatialdistributionoftheaverageannualrainfall.Thepropagationof
quantitativemethodtoestimatestreaminflowintothesmallreser- streamwaterwasthenoperatedfromthemostupstreamreachto
voirs. Three of the main model components are detailed below: thecatchmentoutletbyconsideringthewatervolumeintercepted
i) the small reservoir water balance model, ii) the quantitative ineachreservoir.Thecumulativeimpactofreservoirsisthenclas-
method used to quantify inflow to reservoirs, and iii) the spatial sicallythedifferencebetweensimulateddischargeandthegauged
representation of the reservoir network. The inflow quantification discharge.InTEDI,Nathanetal.(2005)usedtheobserveddischarge
method and the spatial representation of the reservoir have to be oftheperiodofinterest.Theinflowineachreservoiriscalculated
consistent and are thus intrinsically dependent. A spatially dis- fromtheobservedcatchmentdischargeassumingaproportionality
tributedrepresentationofreservoirsrequiresbeingabletoestimate with the reservoir catchment area. The outflow from every reser-
the spatial distribution of stream flow to estimate the upstream voir is transfered directly to the outlet. It is then considered that
inflow to each reservoir. Conversely, an aggregated estimation of theobtainedcumulativeimpactcorrespondstotwicethesimulated
stream flow over a sub-basin or over the full catchment leads to impactofthereservoirnetworkbecausethegaugedischargealready
thereservoirnetworkrepresentationbeingaggregatedonthesame includestheimpactsofexistingreservoirs.
domain.
Mostofthereviewedstudiesfocusedonassessingtheimpactsof 4.2.3. Reservoirspatialrepresentation
reservoirsusedforirrigationorlivestockwateringonstreamflow.In Howthereservoirnetworkisrepresentedfromaspatialperspec-
suchcases,theimpactsarequantifiedbycomparingthecatchment tivevariesfromonemodeltoanother.Thespatialrepresentationof
streamflowsimulationwithandwithoutreservoirs,exceptforthe thereservoirnetworkcanbeclassifiedintothefollowingthreetypes
TEDImodel,aswewillseeinSection4.2.2.Theexceptionstomod- (seeFig.6,Table1).
ellingapproachesdedicatedtostreamflowimpactsarethoseaiming
atassessingtheimpactsongroundwater.Theseapproachesmostly • Inthespatiallyaggregateapproach,allthereservoirsinacatch-
focus on infiltration tanks, for which part of the stored volume ment(inTable1,Aforaggregationonsub-catchmentsandA*
rechargestheaquifer.Insuchcases,onlytheimpactsontheaquifer foraggregationonagridcell)arerepresentedintheformofa
duetothelossfromthereservoirsarerepresented,eitherwithout singleequivalent,orcomposite,reservoir.
simulationofthegroundwater(Martín-Rosalesetal.,2007;Hughes • The statistical representation constitutes a refinement of the
and Mantel, 2010), with a simplified representation of the aquifer aggregate representation (Fig. 6B). The reservoir network is
(Smoutetal.,2010;Shindeetal.,2010;Perrinetal.,2012),oreven represented in the model in an aggregated way by grouping
moreseldom,witha2-Dhydrogeologicalmodel(Ramireddygariet reservoirs into a finite number of classes. Some hydrological
al.,2000;Barberetal.,2009). connections between several of these classes may be repre-
sented(SinTable1).
• The spatially explicit representation consists of representing
4.2.1. Reservoirwaterbalancemodel everyreservoir(Fig.6C).
ReservoirwaterbalancemodelsrelyonEq.(1).Mostsmallreser-
voir water balance models take into account the evaporation and
abstraction,forwhichtemporalestimationisrarelywellknownand 4.2.4. Aggregaterepresentation
isoftenanimportantpoint(seeSection5.3)(Table1).Whenseep- In the aggregate representation (Fig. 6A), the characteristics of
age is taken into account, it is considered only as infiltration to the equivalent reservoir (capacity and surface area) are obtained
groundwater. Ignoring seepage is justified by the small expected byaggregatingsinglereservoircharacteristics.Themaininterestof
rates (Hughes and Mantel, 2010) or by the lack of information on the aggregate representation is to require only global information
theprocess(rate,timing,anddrivingfactorGüntneretal.,2004)and about the reservoirs and their characteristics. In fact, the spatial
bythefactthatseepagefluxcancontributetodownstreamflow.To densityofreservoirswithinacatchmentcanbelarge,greaterthan
simulate the reservoir water mass balance, downstream discharge 10reservoirs/km2insomecases(Nathanetal.,2005),andanexhaus-
issimulatedconsideringthatreservoirsoperatewiththetechnique tive inventory of all reservoirs along with their characteristics is
of “fill-and-spill” (Section 3.3, unless a conservation flow is taken out of reach. Rather, a global estimation of reservoirs and their
into account (Table 1). Reservoir inflow is simulated by different characteristics may be approximated from simple rules of spatial
approaches,aspresentedinthenextsection. extrapolation(cf.Habetsetal.,2014).Forinstance,toestimatethe
858 H.Florenceetal. /ScienceoftheTotalEnvironment643(2018)850–867
Table1
Mainprocessesinreservoirwaterbalancemodel,aswellastemporalandspatialrepresentationsofreservoirsinnumericalmodels.Spatialrepresentationcanbethefollowing
(seeFig.6):A:aggregaterepresentationbycatchment(orA*bygridingrid-basedmodels),S:statisticalrepresentation,orD:distributedrepresentation.Inflowtothereservoirs
canbederivedfromOBS:observations,WB:waterbalance,orEWB:energyandwaterbalance.Outflowiscomputedeitherbasedonspill(aboveawaterlevelorvolumeinthe
reservoir)and/ortakingintoaccountaconservationflow(CF).
Processesincluded
Model Spatialrepresentation Timestep Inflow Outflow Evaporation Directprecipitation Seepage Aquifer Abstraction
ACRUi A Day WB CFspill x x x x
GR4Jk A Day WB
HYDROMEDl A Day WB Spill x x x
POTYLDRj A Day WB CFspill x ? x ? x
ISBA-Rapidh A* Hour EWB Spill x x
SWAT Ag,Sn Day WB Spill x x x x
TEDIa S Month/day OBS Spill x x x
WASAd S Day WB Spill x x x
WaterCASTc S Day WB Spill x x x
CASCADEm D Day WB Spill x x x
CHEATb D Month OBS Spill x x x
Deitchetal.e D Day OBS Spill
PITMANf A Month WB Spill x x
a Nathanetal.(2005).
b Nathanetal.(2005).
c Cetinetal.(2009).
d Güntneretal.(2004).
e Deitchetal.(2013).
f HughesandMantel(2010).
g Perrinetal.(2012).
h (Habetsetal.,2014).
i TarbotonandSchulze(1991).
j Ramireddygarietal.(2000).
k Payanetal.(2008).
l Ragabetal.(2001).
m Shinogietal.(1998),Jayatilakaetal.(2003).
n Zhangetal.(2012).
inflow into the equivalent reservoir, it is necessary to determine onereservoirtoanotherbutquitehomogeneousinreservoirsofsim-
the contributive catchment. It can be a fraction of the catchment ilarsizes.Inthisway,itovercomesoneofthemainshortcomingsof
area (Tarboton and Schulze, 1991; Hughes and Mantel, 2010) that theaggregaterepresentation.Evaporation,forexample,dependson
can be estimated from the sum of the drainage area of all reser- thewatercolumnheightandcirculationwithinthereservoir,which
voirsordependingonthecumulativereservoirarea(Habetsetal., isexpectedtodependonreservoirsize(cf.Section3.2).Connectivity
2014). to the network – reservoirs and rivers – and operation rules may
Theaggregaterepresentationleadstoobtainingasimulationof also be different depending on the reservoir function, which also
thehydrologicalcumulativeimpactsofreservoirsatthecatchment, depends on the reservoir size. Another advantage of the statisti-
grid-cell or sub-catchment outlet but intrinsically does not allow cal representation is being computationally faster than the fully
simulatingthecumulativeimpactsalongtherivernetworkfromthe distributed one because fewer reservoir mass balances have to be
head to the outlet, unless the sub-catchments are small, which is computedandwatertransfersbetweenreservoirsaresimplified.The
often not the case because the size of the sub-catchment is often mainshortcomingisthatitdoesnotobtaindistributedsimulationsof
determined by the availability of river gauges. Furthermore, this thehydrologicalimpactsofreservoirs;particularly,thecumulative
representation may not reflect the different responses of the vari- impactsalongthefullrivernetworkcannotbesimulated.
ous reservoirs in terms of key processes (evaporation, infiltration,
operations,andsoforth;Zhangetal.,2012).
4.2.6. Distributedrepresentation
A distributed representation of the reservoir is the only way
4.2.5. Statisticalrepresentation toexplicitlyrepresenttheinteractionsbetweenreservoirsbycon-
The statistical representation is a trade-off between the other sidering the outflow from one reservoir as a contribution to the
tworepresentations.Itconsidersthatinformationaboutthelocation inflowofthedownstreamoneandtheinteractionsbetweenreser-
andcharacteristicsofreservoirs,particularlyofsmall-andmedium- voirs and hydrological compartments (river, soil, and aquifer) by
sizedreservoirs,cannotbeexhaustivelyavailable.Italsoreliesonthe estimatingtheimpactsofeachreservoironitsconnectedriverreach
assumptionthatreservoirconnectivitymayplayaroleinthecumu- or/and aquifer. Indeed, two dams with similar characteristics may
lative impacts. The reservoir network is represented by classes of havedifferentimpactsaccordingtotheirlocationalongthestream
reservoirsdeterminedfollowingreservoirwatercapacity(Güntner network, mostly because the inflow is not the same. The interest
etal.,2004;Nathanetal.,2005;Loweetal.,2005)andalsoreser- in a spatially explicit representation is in quantifying and under-
voirdrainagearea(Zhangetal.,2012).Eachclassisrepresentedas standingthelocalhydrologicimpactatariverreachscaleandthe
asingleequivalentreservoir.Güntneretal.(2004)andZhangetal. cumulative impacts along the river network (Deitch et al., 2013).
(2012)usedacoupledsequentialandparallelschemetorepresent Quantifying local hydrologic impacts may be particularly relevant
theupstream-downstreamconnectivityofdifferentwaterreservoir towaterquality,ecologicaldisturbanceormorphogenesisevolution.
classesinthecatchment. Inaspatiallyexplicitrepresentation,waterinflowintoeverysingle
Asamainadvantage,thestatisticalrepresentationhastoconsider reservoir as stream discharge and lateral surface runoff has to be
thediversityofkeyreservoirprocesses,whichcanbevariablefrom knownorestimated.Toourknowledge,onlyShinogietal.(1998)and
H.Florenceetal. /ScienceoftheTotalEnvironment643(2018)850–867 859
Fig.6. Spatialrepresentationofreservoirnetworkinmodelsusedtoquantifycumulativereservoirhydrologicimpacts.
Smout et al. (2010) have performed catchment hydrologic mod- of information and the difficulty to obtain it exhaustively, statisti-
ellingtoobtaintheseestimations,withanapplicationtorelatively calrepresentationsandaggregaterepresentationsareconsideredas
simplecasestudiescharacterizedbyfewreservoirs.Otherreported pragmaticalsolutionsandusedinmostmodellingstudies.
casestudiesusingspatiallyexplicitrepresentationsusedobserved-
stream-discharge-based models (Nathan et al., 2005; Cetin et al.,
2009;Deitchetal.,2013). 5. Howtoobtainaccesstotheinformationneededonsmall
A spatially explicit representation relies on the availability of reservoirs?
exhaustive information about reservoir location, characteristics,
water uses and topology, which are rarely available over large 5.1. Whattypeofdata?
areas. This point constitutes a main shortcoming of the approach,
asaddressedinSection4.1.1.Furthermore,itcanbeexpectedthat Stream discharge time series, at one or several points in the
uncertainties in the local information, added to the uncertainty catchment, are required data in statistical analyses (Section 4.1.2)
in estimated spatial discharge and individual reservoir water bal- and in the TEDI and Deitch models (Nathan et al., 2005; Deitch et
ances, can skew the local simulated impacts, and by propagation, al., 2013, Section 4.2.1), and such data are also used by the other
the cumulative impacts. This could alleviate the theoretical inter- types of models to calibrate or assess the modelling. Such data
estinthespatiallyexplicitrepresentation.Acknowledgingthelack areexpectedtobefoundinexistingdatabases.Statisticalanalyses