building joint india-uk capacity, capability, research and ...€¦ · building joint india-uk...
Post on 23-May-2020
1 Views
Preview:
TRANSCRIPT
1
BuildingjointIndia-UKcapacity,capability,researchand
innovationintheenvironment:Finalreport
AmandaJ.Robinson
JulianR.Thompson
MikeAcreman
KatjaLehmann
EgonL.Dumont
NathanJ.Rickards
T.ThomasManishNemaPrabhashMishraShikhaGaurPushpendraAgarwalYatveerSinghSharadJain
2
TableofContents
1.Introduction............................................................................................................................................32.HydrologicalmodellingoftheUpperNarmadaBasin..............................................................52.1.MIKESHEmodellingoftheUpperNarmadaBasin............................................................................52.1.1.Modeldevelopment.................................................................................................................................................62.1.2.Modelcalibrationandvalidation....................................................................................................................142.1.3.Results:Modelcalibrationandvalidation...................................................................................................14
2.2.GWAVAmodellingoftheUpperNarmadaBasin.............................................................................202.2.1.Modeldevelopment..............................................................................................................................................202.2.2.Modelcalibrationandvalidation....................................................................................................................262.2.3.Results:Modelcalibrationandvalidation...................................................................................................27
2.3.HydrologicalmodellingoftheUpperNarmada:Summaryandongoingwork.....................313.GangaMetagenomePilotStudy.....................................................................................................323.1.Methods.........................................................................................................................................................323.1.1.Sampling....................................................................................................................................................................323.1.2.DNAextraction........................................................................................................................................................343.1.3.Sequencing................................................................................................................................................................35
3.2.Preliminaryresults...................................................................................................................................353.3.Summaryandongoingwork..................................................................................................................35
4.Conclusions..........................................................................................................................................365.References............................................................................................................................................36Appendix1.Expenditure.....................................................................................................................42
3
1. Introduction
This project involved collaboration betweenUCL and the Centre of Ecology andHydrology
(CEH) in the UK and the Indian National Institute of Hydrology (NIH), with the aim of
supportingtheimprovementofwater-relatedscienceinIndia.
Therearetwomainstrandstotheproject.Thefirstfocussedonhydrologicalmodellinginthe
UpperNarmadaBasin.Hydrologicalmodellingisakeytoolforimprovingourunderstanding
ofwater resources and assessing the potential impacts of a variety of scenario types upon
waterresources, fromclimate (e.g.Thompson etal.,2013;2014a;Ho etal.,2015)and land
cover change (e.g. Kalantari et al., 2014; Wijesekara et al., 2014), to irrigation and dam
regulation scenarios (e.g. Ahrends et al., 2008; Singh et al., 2011; Räsänen et al., 2012). It
thereforehasanimportantroletoplayincatchmentmanagementandplanning(Loucksand
van Beek, 2005; McCartney and Acreman, 2009). However, uncertainties are inevitably
introduced intoallhydrologicalscenario impactassessments(e.g.RefsgaardandHenriksen,
2004;Goslingetal.,2011),anditisimportantthattheresultantuncertaintyinthemodelling
resultsisrecognisedandtakenintoaccount.
Onesourceofuncertaintythathasbeenrelativelyunder-studiedisinter-hydrologicalmodel
uncertainty (Prudhomme and Davies, 2009; Thompson et al., 2013). There are numerous
hydrological model codes and these vary in terms of their representation of hydrological
processes,spatialdiscretizationanddatarequirements.Inter-hydrologicalmodeluncertainty
arises because although different hydrological models of the same catchment, developed
usingalternativemodel codes,mayallperformacceptably foranobservedbaselineperiod,
they may still respond differently under scenario conditions (e.g. Gosling et al., 2011;
Hagemann etal.,2013;Thompson etal.,2013;Velázquez etal.,2013).Oneobjectiveof this
project was therefore to develop alternative models of the same catchment using two
differentmodelcodes, inordertoenable future inter-modelcomparisons.Morespecifically,
thesemodelswill beused for scenariomodelling aspart of ongoing collaborationbetween
UCL,CEHandNIH,anduseofalternativemodelcodeswillallowamorerobustassessmentof
therangeofpotentialimpacts.ThisongoingworkisdiscussedfurtherinSection2.3.
TheNarmadaBasinwasselectedasasuitablecatchment for theapplicationofhydrological
modelling and subsequent inter-hydrological model uncertainty assessment. With a
populationofover16million(GovernmentofIndiaMinistryofWaterResources,2014)anda
4
drainageareaof98,796km2(India-WRIS,2015),theNarmadaBasinisanexampleofariver
basinfacingnumerousmanagementchallenges.Inparticular,therearemultipleongoingand
planneddamandirrigationdevelopmentprojectsforthebasin(GovernmentofIndiaMinistry
ofWaterResources,2014).Atthesametime,itisvitalthatenvironmentalflowrequirements
(the flow needs of the river ecosystem; Richter et al., 1997; Acreman and Dunbar, 2004),
continuetobemet, inordertosustaintheeconomically,sociallyandecologically important
ecosystemservicesprovidedby theriver. Importantly,NIHwereable tomakeavailable to
UCL and CEH observed river discharge records for multiple sites within the basin. Such
recordsareaprerequisiteforhydrologicalmodelcalibration/validationandaccesstorecords
forsomeriverbasinsinIndia,mostnotablytheGanges,canbeproblematical(Johnstonand
Smakhtin,2014;Masoodetal.,2015).
Thesecondstrandofthisprojectcomprisesapilotstudytoassesstheecologicalstateofthe
upperGangausingametagenomicapproach.OnlyfewmicrobialdatasetsoftheGangaexistto
date, partly because it is difficult to sample eDNA (environmental DNA) under often
unfavourable conditions (e.g. heat, long journey times to sampling points, inaccessibility,
insufficient laboratory facilities) and partly because whole metagenome sequencing is a
relativelyyoungmethod.Thisbio-assessmentwillprovideasnapshotofthecurrentcondition
of the microbial, invertebrate and vertebrate components of the river ecosystem from its
confluencetothefirstsizeableurbancentres.Overandabovethat,itwillprovideacontextfor
spatial and temporal studies inyears to come,which is likely tobeextremelyuseful in the
lightofinitiativessuchastheCleanGangacampaign,launchedbytheIndianGovernmentlast
May. It will assess the potential of eDNA as a cost effective tool to biomonitormicro- and
macrobiota and investigate water quality issues (Tan et al., 2015) and ecological state
(Thomsen et al., 2015) of the river and its tributaries. The pilot study involved four field
campaignstosample13samplingsites(orasubsamplethereof,Figure3.1)fromtheriver’s
confluence at Devprayag past the cities Rishikesh and Haridwar, where discharge of
pollutantsandabstractionofwaterincreases,totheMadhyaGangabarrage,wheretheflowof
theriverisblockedandthenpartlydivertedintoacanalbeforeitenterstheplains.Thiswill
helptoinvestigatethedegreeofecologicaldeteriorationthattheriverexperiencesinitsfirst
100km.
5
2. HydrologicalmodellingoftheUpperNarmadaBasin
TheNarmadaRiverislocatedincentralandwesternIndiaandisthelargestwesternflowing
river of thepeninsula India (Government of IndiaMinistry ofWaterResources, 2014). The
basin largely falls within the State of Madhya Pradesh, but also covers parts of Gujarat,
MaharashtraandChhattisgarh(Figure2.1).Thehydrologicalmodellingstrandoftheproject
focuseson theUpperNarmadaBasindowntoHoshangabad,which lieswithin theStatesof
Madhya Pradesh and Chhattisgarh only. Hydrological models of the Upper Narmada were
developedusingtwodifferentmodelcodes:MIKESHEandGWAVA.Thedevelopmentofthese
models isdiscussedinSections2.1and2.2,respectively.Section2.3providesasummaryof
theoutcomesofthisstrandoftheprojectandanoutlineofongoingwork.
Figure2.1.TheNarmadaBasin(left)andtheUpperNarmadaBasin(right).
2.1. MIKESHEmodellingoftheUpperNarmadaBasin
In consultationwithNIH, aMIKESHEhydrologicalmodel of theUpperNarmadaBasinhas
beendevelopedatUCL,usingapproachesdevelopedfortheMekongRiverBasin(Thompson
et al., 2013; Thompson et al., 2014a; 2014b).MIKE SHE is a comprehensive, deterministic,
distributedmodellingsystem,capableofsimulatingthemajorprocessesofthelandphaseof
thehydrologicalcycle(GrahamandButts,2005). Ithasamodularstructureandalthoughit
was originally designed as physically-basedmodel code,many of themodules now offer a
rangeofprocessdescriptions,someofwhichareconceptualandsemi-distributed.Theseare
6
particularlyapplicableforlargebasinssuchastheNarmadawherethefocusisthesimulation
ofriverflowandwheredetaileddata,suchashydrogeologicalcharacterisation,requiredfor
morephysically-basedapproaches,arenotavailable(Andersenetal.,2001;Stisenetal.,2008;
Refsgaardetal.,2010).
2.1.1. Modeldevelopment
Table2.1summarisesthecomponentsoftheMIKESHEmodeloftheUpperNarmadaandthe
dataemployedwithinthem.Themodeldomainwasspecifiedusingashapefilebasedupona
catchmentshapefileforthewholeoftheNarmadaBasinthatwasprovidedbyNIH.Watershed
delineation was undertaken using topographic data (see below) and this shapefile within
ArcSWAT(Winchelletal.,2013)todefinethecatchmentextentoftheUpperNarmadaBasin
to just downstream of Hoshangabad discharge station. This provided a catchment area of
44,725km2.Themodelgridsizewassetto2km×2kminordertoretainabalancebetween
representingcatchmentcharacteristicsandefficientcomputationtime(Vázquezetal.,2002;
Thompson et al., 2013). Consequently, although most spatial inputs to the model, such as
topographyand landcover,havea resolutionof 1km×1km,duringmodel runs, all input
dataareautomaticallyresampledtothe2km×2kmmodelgrid.Topographywasspecified
usingSRTM(ShuttleRadarTopographyMission)baseddata (Figure2.2;available fromthe
USGSEarthExplorer:http://earthexplorer.usgs.gov/).
Thespatialdistributionoffivedifferentlandcoverclasseswithinthebasinwasobtainedfrom
NIH(Figure2.2).Thesedatawerebasedonremotesensing.LeafAreaIndex(LAI)valuesfor
the different land cover classeswere based on a combination of those used in theMekong
MIKE SHEmodel and values from Jain et al. (1992), a study thatmodelled the Kolar sub-
catchmentoftheNarmadaBasin.RootDepth(RD)valuesforthedifferentlandcoverclasses
werebasedonthoseusedintheMekongMIKESHEmodel(seeTable2.2).Overlandflowin
theUpperNarmadamodel iscalculatedusinga finite-differenceapproach tosolve the two-
dimensional Saint–Venant equations (Graham and Butts, 2005). The land cover data were
employedtospatiallydistributeManning’sM(theinverseofManning’sn)valuesforoverland
flowresistance(Table2.2),withvaluesbasedonVieux(2004).
7
Table 2.1. Summary of key data (and data processing) required for each component of the
coupledMIKESHE/MIKE11modeloftheNarmada.
Modelcomponent
Keyinputs/datarequired
Initialdataformat ProcessingundertakenexternallytoMIKEZero*
ProcessingundertakeninMIKEZero
Modeldomain Catchmentextent–thebasinareaupstreamofHoshangabad
ESRIpolygonshapefile
InArcGIS:Convertedtoadifferentcoordinatesystem(ProjectedcoordinatesystemWGS1984UTMZone44N).EditedtheedgetopreventgriddingerrorsinMIKESHE.
N/A
Topography Topography ESRIgridrasterfilewitharesolutionof0.0008333°×0.0008333°.Datasource:SRTM(ShuttleRadarTopographyMission)
InArcGIS:ConvertedtoUTM44Ncoordinatesystemandtoaspatialresolutionof1km×1kmtomatchtheinitialmodelgrid.Finalmodelgridsize:2km×2km.
ConvertedtoMIKEZero’sdfs2gridfileformat.
Landuse/vegetation
Landusedistribution
Imagineimagerasterfile.Therearefivelanduseclasses:Forest,Shrub,Waterbodies,BaresoilandAgriculture.
InArcGIS:ConvertedtoUTM44Ncoordinatesystemandtoaspatialresolutionof1km×1km.Buffergeneratedaroundavailabledata.
ConvertedtoMIKEZero’sdfs2gridfileformat.
LeafAreaIndexes
N/A ValuesforthedifferentlandcoverclassesarebasedonthoseusedinsimilarworkundertakenbyUCL(Thompsonetal.,2013;Thompsonetal.,2014).
Rootdepths N/A Asabove.
Overlandflow:modelledusingthe2Dfinite-differencemethod
Manning’sMforoverlandflowresistance
N/A Manning’sMvaluesdistributedaccordingtolanduse/landcover,usingadfs2gridfilebasedonthelandusedfs2file.ValuesforthedifferentlandcoverclassesbasedonVieux(2004).
Unsaturatedzone:modelledusingthetwo-layerwaterbalancemethod
Soilclasses ArcGISpolygonshapefileandrastersthatprovidesoilclassdataforthemajority(~80%)ofthebasin.
InArcGIS:ConvertedtoUTM44Ncoordinatesystemandtoaspatialresolutionof1km×1km.Buffergenerated.
ConvertedtoMIKEZero’sdfs2gridfileformat.
Soilhydraulicproperties
N/A Valuesforthedifferentsoilclassesderivedfromtheliterature(ClappandHornberger,1978;NormanandDixon,1995).
8
Table2.1.(cont.)
Modelcomponent
Keyinputs/datarequired
Dataobtained/ReceivedfromNIH–Dataformat
ProcessingundertakenexternallytoMIKEZero
ProcessingundertakeninMIKEZero
Saturatedzone:modelledusingtheconceptual,linearreservoirmethod
Spatialdistributionofground-watersub-catchments,interflowreservoirsandbaseflowreservoirs
NIHconsultedwithonthenumberandlocationofgaugingstationstobeemployedduringmodelcalibration.
PolygonshapefilehasbeengeneratedinArcGIS.Thebasinwasdividedintofivegroundwatersub-catchmentsbasedontopographyandthelocationsofthefivecalibrationgaugingstations.
N/A
Catchmentmeteorology:Precipitationandevapo-transpirationmodules.
Spatialdistributionofprecipitation
N/A Polygonshapefileof0.25°×0.25°gridsquarescoveringthebasinextent,generatedinArcGIS.
N/A
Precipitation 0.25°×0.25°griddeddailyprecipitationforIndiafortheperiod1901–2013.Dataforeachyearisinadifferentfile,inagriddedbinaryformat.
DatafirstconvertedtoASCIIfileformatusingCscriptprovidedwithdata.DatathenprocessedinMatlabintothenecessaryformatforinputtoMIKEZero.
DataconvertedintoMIKEZero’sdfs0timeseriesfileformat.
Spatialdist.ofPET
N/A Generated1°×1°polygonshapefile.
N/A
Potentialevapotrans-piration
Gridded(1°×1°)temperaturedatareceivedfromNIHforthecalculationofPET.
PETdatacalculatedinMatlab–forthecellscoveringthebasindowntoHoshangabadonly.
DataconvertedintoMIKE’sdfs0timeseriesfileformat.
MIKE11one-dimensionalhydraulicmodelforsimulatingchannelflow(usingKinematicrouting)
Planofthemainriverchannels
AnESRIpolylineshapefile.Thisdenserivernetworkisdividedintosevenstreamorders.
InArcGIS:Differentstreamordersextractedtoseparateshapefiles.Newfieldgenerated:lengthofeachbranch.
RivernetworkdigitisedinMIKEZero/MIKE11.
Syntheticcross-sections
N/A Syntheticcross-sectionsgeneratedinExcel.
DatainputtoMIKE11model.
Manning’snforbedresistance
N/A Representativevaluebasedontheliterature(Chow,1959)andpreviousmodellingexperience.
Reservoirregulationdata/rules
NIHwillbeconsultedastowhetherreservoir/damregulationdatacanbemadeavailableforfuturemodeldevelopment.Currently,modelcalibrationhasbeenundertakenwithoutdamsincludedinthemodel.
N/A Riverdischargetimeseriesformodelcalibration/validation.
Dataforthecalibration/validationperiodof2002–2013receivedforfivegaugingstations.
DataforthefivestationshavebeenprocessedinExcelandMatlabintothenecessaryformatforinputtoMIKEZero.
DataconvertedintoMIKEZero’sdfs0timeseriesfileformat.
9
Figure2.2.TopographicdatafortheUpperNarmadaBasin.
Figure2.3.LandcoverclasseswithintheUpperNarmadaBasin.
Table2.2.Summaryoflandcoverclassesandrelatedparameters.
Code Landcover Rootdepth(mm)1 LAI2 Manning’sM32 Forest 900 1.0–6.0 103 Shrub 500 2.6–3.8 84 Waterbodies 0 0 255 Baresoil 10 0 506 Agriculture 500 0.3–6.9 29
1BasedontheMIKESHEmodeloftheNarmada.2BasedontheMIKESHEmodeloftheNarmadaandJainetal.(1992).3BasedonVieux(2004).
Land cover
AgricultureBare SoilForestShrubWater Bodies
maslHigh : 1317
Low : 279
0 10050 km¯
0 10050 km¯
Land cover
AgricultureBare SoilForestShrubWater Bodies
maslHigh : 1317
Low : 279
0 10050 km¯
0 10050 km¯
10
Following initial calibration attempts that revealed overestimation of discharge at
downstream discharge stations (Barmanghat and Hoshangabad), irrigation was included
within the model over two command areas: Bargi (1570 km2) and Barna (579 km2). The
locationsofthecommandareas(seeFigure2.4)werebasedonafigurefromGovernmentof
IndiaMinistryofWaterResources(2014)thatwasgeorectifiedanddigitisedinArcGIS.Data
on the location of the culturable command area (land actually irrigated) within the gross
commandarea(theoverallregioncontainingirrigatedland)werenotavailable.However,the
acreagesofthecommandareasincludedintheMIKESHEmodelweremadetomatchthose
reported on the India-WRIS (Water Resources Information System) website (India-WRIS,
2013a,b)and inGovernmentof IndiaMinistryofWaterResources(2014). Irrigationwater
for the Bargi and Barna command area was specified as being abstracted from the river
sectionsatthelocationsofBargiDamandBarnaDam,respectively.Duringmodelcalibration,
anevapotranspirationcropcoefficient(Kc)of1.2wasspecifiedoverthecommandareasfor
themonthsofMay–September.ThismeansthattheinputPETovertheseareasismultiplied
by 1.2 in these months. Crop coefficients are commonly employed to adjust potential
evapotranspirationestimatesspecificallyforcropland,andaKcof1.2iswithintherangeof
normalKcvaluesaccordingtoAllenetal.(1998).
The two-layer water balance method was employed for the unsaturated zone. For this
module,thespatialdistributionofsoilclasseswasspecifiedusinga1km×1kmgridbasedon
a georectified and digitised version of a soil map that was provided by NIH. Soils were
aggregated intosixclasses (Figure2.5)and therequiredhydraulicparameters foreachsoil
classweretakenfromtheliterature(Table2.3).
Figure2.4.GrosscommandareasandassociateddamsincludedinMIKESHE.
!
!
Barna DamBargi Dam
Soil classClayClay loamLithosol
SandSandy loamSilt loam
Bargi command area
Barna command area
0 10050 km¯
0 10050 km¯
11
Figure2.5.SoilclassdistributionwithintheNarmadaBasin.
Table2.3.Soilclassesandassociatedparametervalues.
Soil_code
Soil_class Watercontentatsaturation*
Watercontentatfieldcapacity+
Watercontentatwiltingpoint+
Saturatedhydraulicconductivity(m/s)
1 sand 0.40 0.14 0.04 0.000182 sandyloam 0.43 0.26 0.09 3.5e-0053 siltloam 0.49 0.34 0.16 7.2e-0064 clayloam 0.47 0.34 0.18 2.5e-0065 clay 0.48 0.42 0.25 1.3e-0066 lithosol 0.45 0.30 0.13 7e-006
*fromClappandHornberger(1978) +fromNormanandDixon(1995)
Formodelling thesaturatedzone, theconceptual, semi-distributed, linear reservoirmethod
was selected. Advantages of this method include lower data requirements and reduced
computationtimecomparedtophysicallybasedsolutions(Andersenetal.,2001;Stisenetal.,
2008;Thompson etal.,2013;2014a;2014b).Using theapproachemployed for theMekong
(Thompson et al., 2013), theUpperNarmadawas divided into five sub-catchments (Figure
2.7)basedupontopographyandthelocationofthedischargegaugingstationsforwhichdata
for model calibration/validation were available. Within each sub-catchment, the saturated
zoneisrepresentedbyashallowinterflowreservoir,andtwobaseflowreservoirstosimulate
fasterandslowerbaseflowstorage.Exchangesbetweenreservoirs,andultimatelytheMIKE
11hydraulicmodel,arecontrolledbytimeconstants(DHI-WE,2009).Thetwotimeconstants
(interflowandpercolation) for each interflowreservoir and thebaseflow time constant for
eachbaseflowreservoirwerevariedduringmodelcalibration.
!
!
Barna DamBargi Dam
Soil classClayClay loamLithosol
SandSandy loamSilt loam
Bargi command area
Barna command area
0 10050 km¯
0 10050 km¯
12
Figure 2.6. Conceptual structure of the sub-catchment based linear reservoir saturated zone
module.Source:GrahamandButts(2005).
Figure2.7.Sub-catchmentdistributionandriverdischargegaugingstationlocations.
##
##
#
Manot
DindoriGadarwaraBarmanghat
Hoshangabad
# Discharge stations
River network
Sub-catchments
1 Narmada to Dindori2 Narmada to Manot3 Narmada to Barmanghat4 Shakkar to Gadarwara5 Narmada to Hoshangabad
1
2
3
45
0 10050 km¯
13
Daily griddedprecipitation data for theUpperNarmadawere derived from the IMD (India
MeteorologicalDepartment)/NCC(NationalClimateCentre)HighSpatialResolution(0.25°×
0.25°)LongPeriod(1901–2013)DailyGriddedRainfallDataSetOverIndia(Paietal.,2014),
obtained from the IMD.Duringmodel calibration (see Sections 2.2.2–2.2.3), a precipitation
lapse ratewas introduced over the spatial extent of sub-catchments 1, 2 and 4,which are
upstream sub-catchments located at higher elevations. The three lapse rates were then
subjecttocalibration.Forthecalculationofdailygriddedpotentialevapotranspiration(PET),
IMD/NCChigh resolution (1°×1°) griddeddaily temperaturedata (Srivastava et al., 2009)
wereused.PETwascalculatedusingtheHargreavesmethod,asthismethodisrecommended
by the FAO for use in situations where there are insufficient data to calculate Penman-
Monetith (Allen et al., 1998).Parameters for the equationwerebasedupon thoseobtained
from ECALTOOL, a computer program that provides calibrated values for the CH and EH
parametersoftheHargreavesequation(Pateletal.,2014).Theparametersvarythroughthe
year, with the values for somemonths subjected to further calibration independent of the
ECALTOOL.ThespatialdistributionsofprecipitationandPETinputsareshowninFigure2.8.
Figure2.8.a)Spatialdistributionofa)precipitationinputsandb)PETinputs.
Forthesimulationofchannelflow,MIKESHEisdynamicallycoupledtoMIKE11(Havnøetal.,
1995),aone-dimensionalhydraulicmodel.Aplanofthemainrivernetworkwasdigitisedin
MIKE11.For thegenerationofsyntheticcross-sections,channelwidthmeasurementswere
taken from satellite imagery in Google Earth. A generalised cross-section profile and a
relationshipbetweenchannelwidthandmaximumchanneldepthwerebasedonlimiteddata
available from NIH (a single cross-section for the river channel at Hoshangabad discharge
station)andtheliterature(Rajaguruetal.,1995;Payasi,2015).Cross-sectionswerespecified
as depths relative to the bank, with bank elevations taken from the SRTM DEM (digital
elevationmodel).Thekinematicroutingmethodwasemployed.
a) b)
14
2.1.2. Modelcalibrationandvalidation
Modelcalibrationwasundertakenagainstdischargerecordsfromfivegaugingstationsforthe
period 2002–2008. The calibration parameters were the time constants of the saturated
zone’sinterflowandbaseflowlinearreservoirsandtheprecipitationlapseratesoverselected
sub-catchments(seebelow).Asdescribedabove,irrigationcommandareaswereaddedtothe
modelintheearlystagesofcalibrationandaKcof1.2wasspecifiedovertheseareasforthe
monthsofMay–September.Modelperformanceateachdischargestationwasevaluatedboth
qualitatively,throughvisualcomparisonofgraphsofobservedandsimulateddischarge,and
quantitatively, using model performance statistics. The indicators used were the Nash–
Sutcliffecoefficient(NSE;NashandSutcliffe,1970),thePearsoncorrelationcoefficient(r)and
the percentage deviation in simulated mean flow from the observed mean flow (Dv;
Henriksenetal.,2003).NSEcanvarybetween-1and1,whilstrcanvarybetween0and1;in
both cases, the closer the value to 1, the better the model performance according to that
criteria.InthecaseofDv,thecloserthevalueto0,thebetter.Modelperformanceaccordingto
the NSE and Dv values was classified using the scheme of Henriksen et al. (2008). Model
validationwas subsequently undertaken for the period 2009 toMay 2013 using the same
stationsandperformancestatistics.
2.1.3. Results:Modelcalibrationandvalidation
Table2.4summarisestheoptimisedvaluesofthecalibrationparameters.Precipitationlapse
rates were employed over sub-catchments 1, 2 and 4, following initial model runs that
displayedconsistentunderestimationofdischargeat gauging stationsdownstreamof these
sub-catchments(Dindori,ManotandGadarwara,respectively).Furthermore,thesearethree
upstream sub-catchments that are located at higher elevations and exhibit relatively large
rangesinelevation.Raingaugenetworksinmountainousregionsoftendisplayabiastowards
stationsbeing locatedat lowerelevations,which can lead to systematicunderestimationof
precipitation(e.g.FreiandSchär,1998;Frei etal.,2003;Lietal.,2016).Precipitation lapse
ratescanbeemployedtotryandaddressthisissue(e.g.Immerzeeletal.,2012b;Wijesekara
etal.,2012).Thefinallapseratevaluesarewithintherangeofthosepreviouslyreportedin
mountainousregions(e.g.Immerzeeletal.,2012a;2012b).
15
Table2.4.Finalcalibrationparametervalues.
Sub-catchmentnumber 1 2 3 4 5Sub-catchmentname Din Man Barm Gad HoshPrecipitationlapserate(%/100m) 6 6 7 Interflowtimeconstantforinterflowreservoir 4 6 14 4 14Percolationtimeconstantforinterflowreservoir 4 14 14 4 14Timeconstantforbaseflowreservoir1(days) 35 35 65 65 65Timeconstantforbaseflowreservoir2(days) 250 200 1500 120 350
Model performance statistics for the calibration period are provided in Table 2.5. As
indicated, a shorter period of 2001–2006was employed atManot, due to data availability.
Observed and simulated daily, monthly and mean monthly discharges are presented in
Figures2.9,2.10and2.11,respectively.Theannualriverregimeisrepresentedfairlywellby
themodel, as aremonthly discharges, with good sequencing of the annualmonsoon flood
pulseachievedatalldischargestations.
Table 2.5.Model performance statistics for the calibration and validation periods (validation
shaded).ModelperformanceindicatorsaretakenfromHenriksenetal.(2008).
Station Period Dv DailyNSE
Dailyr
MonthlyNSE
Monthlyr
Dindori Cal:01/02–12/08 -10.22 *** 0.40 ** 0.64 0.83 **** 0.91
Val:01/09–05/08 0.73 ***** 0.58 *** 0.79 0.84 **** 0.93
Manot Cal:01/02–12/06 -8.98 **** 0.53 *** 0.73 0.93 ***** 0.97
Barmanghat Cal:01/02–12/08 3.89 ***** 0.60 *** 0.80 0.82 **** 0.93
Val:01/09–05/10,06/11–05/13 10.39 *** 0.64 *** 0.82 0.79 **** 0.92
Gadarwara Cal:01/02–12/08 -6.77 **** 0.35 ** 0.59 0.64 *** 0.80
Val:01/09–05/10,06/12–05/13 -21.14 ** 0.76 **** 0.89 0.86 ***** 0.97
Hoshangabad Cal:01/02–12/08 6.66 **** 0.63 *** 0.82 0.84 **** 0.95
Val:01/09–05/08 23.16 ** 0.66 **** 0.84 0.77 **** 0.94
Performanceindicator
Excellent*****
Verygood****
Fair***
Poor**
Verypoor*
Dv <5% 5–10% 10–20% 20–40% >40%
NSE >0.85 0.65–0.85 0.50–0.65 0.20–0.50 <0.20
16
Figure 2.9. Observed and simulated daily discharge for the calibration and validation periods
(separatedbydashedline).
17
Figure2.10.Observedandsimulatedmonthlymeandischargeforthecalibrationandvalidation
periods(separatedbydashedline).
18
Figure2.11.Observedandsimulatedriverregimesfortheperiod2002–2008.
The simulated river regimes at Barmanghat andHoshangabad (Figure 2.11) display a bias
towards underestimation of dry season discharges and overestimation of peak monthly
discharges, although year-to-year variation in the pattern and magnitude of monthly
discharges iswellreproduced(Figure2.10).Observeddryseasonflowsmaybehigherthan
thosesimulatedduetodryseasonreleases fromdamsintheUpperNarmadaBasinsuchas
the Bargi, Barna and Tawa Dams. Similarly, observed wet season discharges at a monthly
resolutionmay be lower than those simulated due to these dams. Dams are not currently
includedintheMIKESHEmodelduetoalackofdataondamregulationandreleases.Future
workwillseektoexplorethepotentialofincludingthemajordamsusingtherangeofcontrol
structureapproachesthatareavailablewithinMIKE11,withdamoperationbeingvariedto
improvemodelperformance in theabsenceofdetailed recordsof actualoperation.Despite
these issues,meandischarges arewell representedby themodel,withDv classed as “very
good”to“excellent”atallstations(Table2.5).
19
Model performance at a daily resolution is notably weaker compared to at a monthly
resolution,asdemonstratedinFigure2.9andTable2.5.Usingmonthlydischarges,NSEvalues
forthecalibrationperiodareclassedas“fair”to“excellent”andrvaluesof0.80andaboveare
achieved.Incomparison,atadailyresolution,NSEisclassedas“poor”(fourstations)to“fair”
(atHoshangabadonly, themostdownstreamstation)and r values rangebetween0.59and
0.82.Thisweakerperformanceatadailyresolutionmaypartlyberelatedtothequalityand
spatialresolutionofthegriddedprecipitationandPETdata.
Model validationwas undertaken using the period 2009 toMay 2013. However, observed
dischargerecordswereunavailableforthestationatManot,anddatawereonlyavailablefor
threeandahalfyearsatBarmanghatandtwoandahalfyearsatGadarwara,asindicatedin
Table 2.5 and demonstrated visually in Figure 2.10. As for the calibration period, good
sequencing of the annual monsoon flood pulse is achieved at all fours stations for the
validationperiod(Figure2.10).Furthermore,dailyriscloseto0.8orhigherandmonthlyris
over 0.9 at all stations (Table 2.5), representing a strong positive correlation between
observedandsimulateddischarges.
AtDindori,themodelperformancestatisticsforthevalidationperiodarebetterthanduring
calibration.ForexampleDv is classedas “verygood”, rather than “fair” (aspreviously)and
dailyNSE is classed as “fair”, instead of “poor”. At Barmanghat,Dv displays an in increase,
representing greater overestimation ofmean discharge for the validation period.However,
the daily NSE value indicates an improvement inmodel performance at a daily resolution,
whilstthemonthlyNSEcontinuestobeclassedas“verygood”despiteasmallreduction.At
Gadarwara,althoughmeandischargeshowsgreaterunderestimationforthevalidationperiod
(amorenegativeDv),NSEand r indicateanoverall improvement inmodelperformanceat
thisstation,atbothadailyandmonthlyresolution.Finally,atHoshangabad,Dvdisplaysan
increase, leading to its classification falling from “very good” to “poor”. Despite this,
performanceatadailyresolutionaccordingtoNSEimprovesfrom“fair”to“verygood”,and
themonthlyNSEvalueremains“verygood”.AtBarmanghatandHoshangabad,thetendency
towards overestimation of peak monthly discharges and underestimation of dry season
monthly discharges that was exhibited during the calibration period is repeated for the
validationperiod(Figure2.10).Thisis,again,likelytobeduetotheeffectsoflargedams.
20
Overall, performance of themodel is considered appropriate to allow use of themodel in
furtherinvestigations,suchasclimatechangescenariosimulation(e.g.Thompsonetal.,2013)
andtheassessmentoftheimpactsofclimatechangeuponenvironmentalflows(Thompsonet
al., 2014b), particularly as comparisons between baseline and scenario discharges would
typically bemade at a temporal resolution lower thandaily, such asmonthly or annual. In
addition, and as discussed above, the potential for including dams within the model to
enhancemodelperformancecanbeexplored.
2.2. GWAVAmodellingoftheUpperNarmadaBasin
2.2.1. Modeldevelopment
In consultationwithCEH (with financial supportprovidedby funds external to the current
project),aGWAVAmodeloftheNarmadabasinhasbeendevelopedatNIH.Thisprocesswas
startedwith a 3-day GWAVA training given by CEH staff at NIH Roorkee (India) inMarch
2015.Thiswasfollowedbyan8-dayGWAVAmodellingworkshopinCEHWallingford(UK)in
June 2015 during which CEH gave further guidance to NIH staff. Between and after these
meetings,GWAVAwassetupandcalibratedbyNIHwithcontinuedsupportfromCEH.
Inputdatahasbeenpreparedfortheentirebasin,althoughmodellingiscurrentlylimitedto
theupperbasin.TheGlobalWaterAVailabilityAssessmentmodel(GWAVA)wasdevelopedby
CEHandtheBritishGeologicalSurveyinordertoprovideanimprovedmethodologyforthe
assessmentofwaterresourcesattheglobalscale.Themodelhasthecapabilitytoincorporate
additionalwater resourcecomponentssuchas reservoirs, abstractions,andwater transfers
thatmodifywater quantity and flow regime.Itwasdevelopedwith funding fromDFID (UK
Department for InternationalDevelopment).Themodelwas initiallydevelopedandapplied
onagridwitharesolutionof0.5°latitude×0.5°longitude,butitcannowbeappliedatfiner
resolutions.Thechoiceofgridsizeisacompromisebetweentheneedtorepresentthespatial
variabilityandtheavailabilityofsuitabledata.
GWAVAprovidesacomparisonofwateravailabilityanddemandatthescaleofthegridcell
for the comprehensive assessment of the spatial and temporal variability of the water
resources over a basin. Model outputs include simulated daily flows and a cell-by-cell
comparison of water availability. The model is also capable of carrying out the combined
21
assessmentofbothsurfacewaterandgroundwater.Policymakersandstakeholderscanuse
model results for taking appropriate resource allocation decisions. GWAVA also has the
capabilitytoassessfuturewaterresourcesunderprojectedchangesindriverssuchasclimate
andlanduse.ThemodelisaFORTRAN90programandrunsonaPCunderWindows7,ina
MS-DOS box. GWAVA is based on the PDM (Probability Distributed Model) rainfall-runoff
model(Moore,1985,2007)anditsconceptualstructureisgiveninFigure2.12.
Figure 2.12. Conceptual structure of the PDM-based GWAVA Rainfall-Runoffmodule. Adapted
from:Moore(2007).
Themodel iscomprisedofthreecomponents: i)pre-processor, ii)coreengineandiii)post-
processor.Asthemodelinputsarerequiredinbinaryformat,thepre-processorconvertsthe
ASCII input files intobinary formatandalso simultaneously checks forerroneousvalues in
thedatasets.Thecoreengineisthemaincomponentofthemodel.Itreadsthepre-processed
filesandrunsthewaterbalancemodel.Thecoreenginealsogeneratesthewateravailability
indices and can be run both in normalmode and in calibrationmode. The post processor
interpretsthedataproducedbythecoreengineandgivesoutputinthedesiredformats.
Thedatarequirement for theGWAVAmodelapplication includesadrainagemapandbasin
boundary, Digital Elevation Model (DEM), Land use/Land cover (LULC) map, soil map,
22
reservoir command area maps, crop type map in command areas and population and
livestock maps. The required climate inputs include daily temperature, precipitation and
potentialevapotranspiration(PET).Otherpotentialinputs,dependingonthewaterresource
components includedwithin themodel, are reservoir area-elevation-capacity data, average
water consumption rate forhumanand livestockpopulation, industrialdemands, irrigation
demands and inter-basinwater transfers. Themodel is capable of simulating the effects of
natural features suchas lakes,wetlands, glaciers and snowalongwith the incorporationof
reservoirs and their impacts on the water availability scenario. Based on these data
requirementsandtheobjectivesof themodellingexercise,multiple input filesneededtobe
preparedintheformatdesiredbythepre-processoroftheGWAVAmodel.Twomajortypesof
inputfilesinclude,i)compulsoryinputfilesandii)optionalinputfiles.Compulsoryinputfiles
include the physical parameter file, general water demands file and the climate data file,
whereas optional files include the sub-catchment calibration file, mountain region
information, glacier information, groundwater information, daily water demands, water
transfersandclimatescenarios.
The physical parameter file contains the information pertaining to area of grid cell, flow
directioninthegridcell,soilclass,landcover,gridcoordinates,channel-routing,percentage
of lake/wetland and lake characteristics such as surface area capacity, shape parameter,
outflow constant and value of the outflow power. The soil and land-use data are used to
determinethemajorparametersofthemodel.AspertherequirementsoftheGWAVAmodel,
thesoilmaphasbeenreclassifiedintosevenmajorclasses:sand,sandyloam,silt loam,clay
loam,clay,lithosolandorganic.ThesoilmapofthestudyareaisgiveninFigure2.13.Thearea
underthevarioussoilclassesisgiveninTable2.6.Similarly,theGWAVAmodelconsidersthe
effectsofonlyfourtypesoflandusecategories,thesebeingforest,shrub,grassandbaresoil.
Although, for the simulationof abstractions for irrigation,detailed informationon irrigated
crop-typesisused.TheLULCmapofthestudyareahasbeenreclassifiedaccordinglyintofour
majorclassesandisgiveninFigure2.14.TheareaunderthevariousLULCclassesisgivenin
Table2.7.
The SRTM (Shuttle Radar TopographyMission) Digital ElevationModel (DEM) data,which
wasalsoemployedintheMIKESHEmodeloftheNarmada,hasbeenresampledatthedesired
resolutionofGWAVAmodelat0.125°latitude´0.125°longitude.Theflowdirectionmaphas
beenpreparedfromtheprocessedDEMandisgiveninFigure2.15.Thephysicalparameter
23
filefortheNarmadabasinhasbeenpreparedbasedontheinformationextractedfromthese
GISfilesandotherparametervalueswereassignedsuitedtothestudyarea.Themodeluses
the Muskingum method of routing the channel flows between the cells and has two
parameters ‘x’ and ‘k’. The parameter ‘x’ is fixed at 0.5 whereas the parameter ‘k’, which
representsthetimedelayindays,canbeassigned.
Figure2.13.SoilmapofNarmadabasininMadhyaPradesh.
Table2.6.Areaundervarioussoilclasses
Soiltype Area(km2) Area(%)Sand 25333 28.4Sandyloam 5650 6.3Siltloam 13486 15.1Clayloam 182 0.2Clay 44469 49.9
Thewaterdemandfilecontainsthegridcellinformationpertainingtototalpopulation,urban
population,cattlepopulation,populationofsheepandgoats,urbanwaterdemandrate,and
ruralwaterdemandrate,intheappropriateunitsrecognisedbythemodel.Theinformation
pertaining to crop type, crop area, planting month etc., is also given in the general water
demandsfile.Themodelhasthecapabilityofincludingeighttypesofcropsforanygridcell.
24
Crop water demands are calculated based on the crop details and PET for the grid cell.
Thereafterthecropwaterdemandsofall thecropsaresummedupforeachcell.Thewater
demand file has also been prepared based on information extracted from theGIS files and
generalinformationpertainingtothecropcharacteristicsinthestudyarea.
Figure2.14.LanduselandcovermapoftheNarmadabasin
Table2.7.AreaundervariousLULCclasses
LULCclasses Area(km2) Area(%)Forest 20551 21.5Shrub 22643 23.7Waterbodies 1523 1.6Baresoil 3008 3.1Agriculture 47802 50.0
TheclimatedatafiletobeusedintheGWAVAmodelcanbeoftwotypes,i)monthlyclimate
normalsandanomaliesorii)dailyclimate.Thedailyclimatedataneedstobeinbinaryformat
whereasthemonthlyclimatedatacanbeeitherinASCIIorbinaryformat.AswiththeMIKE
SHEmodel of theNarmada, the high resolution gridded precipitation data set prepared by
IndiaMeteorologicalDepartment (IMD)of0.25° latitude´0.25° longitudehasbeenused to
extracttheprecipitationovertheNarmadabasin.Similarly,thegriddedtemperaturedataof
25
IMDataresolutionof1°latitude´1°longitudehasbeenusedtoextractthetemperatureover
Narmadabasin.PEThasbeencomputedusingtheThornthwaitemethod.Thereafterthedaily
climatedatafilehasbeenpreparedinthebinaryformat.
Figure2.15.FlowdirectionmapuptoHoshangabadoftheNarmadabasin.
The GWAVAmodelwas run at a resolution of 0.125° latitude´ 0.125° longitude to have a
detailed understanding of the spatial variability of the water availability in the basin. The
Narmadabasinup to its confluencewith theArabianSeacomprisesof661gridcellsat the
selectedresolution.Allthemandatoryaswellastheoptionalinputfileshavebeenprepared
at this resolutionby resampling theoriginaldatasets.Even though the inputdatahasbeen
preparedforthecompletebasin, themodellingexercisewas initiallybe limitedtoNarmada
upto theHoshangabadgaugingsite.TheNarmadaBasinup toHoshangabadcontains three
major dams: Bargi multipurpose dam, Tawa dam and Barna dam. Also manymore major,
mediumandminorprojectsarebeingplanned in thispartof thebasin. Subsequently, after
gaining enough confidence in the model, the modelling exercise will be extended for the
completeNarmadaBasin,comprisingofmanymorewaterresourceprojects.
26
2.2.2. Modelcalibrationandvalidation
TheGWAVAmodelcanberun inthenormalmodeor inthecalibrationmode.Theoptional
sub-catchmentcalibrationfilecanbecreatedtoenabletheuseofcalibratedparametersinthe
user-selectedsub-catchments.Someofthecriticalmodelparametersthatneedtobegivenfor
the various sub-catchments in the sub-catchment calibration parameter file include the
following:thePDMparameterofpowerlawprobabilitydistribution(b)whichdescribesthe
spatialvariations in soilmoisture storagecapacity, surface runoff routingparameter (Srout),
groundwaterroutingparameter(Grout),multipliertoadjustrootingdepths(fact).
Forthepresentstudy,themodelhasbeenruninthenormalmodetounderstandtheabilityof
themodelinsimulatingtheflowsinthebasinbasedonthemanualcalibrationbyfinetuning
theparametersofthemodelwithintheprescribedparameterranges.Themanualcalibration
has been carried out with the compulsory input datasets andmodel simulated flows have
beencompared toobserved flowsat theHoshangabadgauging site. Several trial runswere
performedand themodel resultshavebeen comparedwith thegauged flowseries and the
modelparametersadjustedaccordinglytoobtainbetterresults.Thelikelyrangesofthemajor
parametersaregiveninTable2.8.
Table2.8.Likelyrangeofthekeyparameters.
Parameter upperbound expectedvalue lowerboundb 4 1 0.25fact 4 1 0.25Srout 12 - 0Grout1 100
1:AnegativevalueofGroutisallowed.However,thismeansthatthereisnoroutingdelay,i.e.allwater
thatdrainsfromthesoilstorebecomesbaseflowimmediatelywithoutbeingtemporarilystoredinthe
groundwaterstore.
2:Sroutvaluesabove1areallowed,howeverthensurfaceroutingismodelledasifSroutis1.
Themodelwasinitiallyrunfortheperiod2001–2006andthemodelparameterscalibrated
using themanual calibration procedure. The post-processorwas run subsequently and the
localflowsaswellasthetotalflowswereobtainedforeachgridcellinthebasin.Themodel
performancehas been evaluated through visual comparisonof the observed and simulated
discharge plots at the grid cell pertaining to the Hoshangabad gauging site. Model
performancehasalsobeenevaluatedusingtheindicatorofpercentagedeviationinsimulated
27
flow from the observed flow at various time steps (Dv; Henriksen et al., 2003). Model
validationwassubsequentlyundertakenfortheperiod2007–2010.
2.2.3. Results:Modelcalibrationandvalidation
Table 2.9 summarises the optimised values of the calibrated parameters obtained by fine
tuning the sensitiveparametersusing themanual calibrationapproach. Comparisonof the
observedandsimulateddailystreamflowatHoshangabadgaugingsiteduringthecalibration
isgiveninFigure2.16.Similarly,comparisonoftheobservedandsimulatedmonthlystream
flowatHoshangabadgaugingsiteduringthecalibrationisgiveninFigure2.17.Thesegraphs
show that themodel has been able to simulate the flows at themonthly scale with a fair
degreeofaccuracycomparedtothedailyflowvalues.Themodelisabletosimulatethepeaks
aswell as the recession curveof thedailyhydrographs reasonablywellwith theminimum
datainputsusingmanualcalibrationbyconsideringthestudyareaasasinglesub-catchment.
Table2.9.Parametervaluesobtainedduringmanualcalibration.
Parameter Calibratedparametervaluesb 0.0065Srout 0.08Grout 0.25bfpower 4bfpower:baseflowrecessionpower
Figure2.16.Comparisonoftheobservedandsimulateddailyflowduringcalibrationat
Hoshangabad.
28
Figure2.17.Comparisonoftheobservedandsimulatedmonthlyflowduringcalibrationat
Hoshangabad.
Theperformanceof themodelhasbeenevaluatedbasedonthepercentagedeviation inthe
simulatedmonthlyflowsfromtheobserved,theresultsofwhicharegiveninTable2.10.The
comparisonofthemeanmonthlystreamflowduringcalibrationperiod2001–2006isgivenin
Figure2.18.
Table2.10.Differenceinvolumeintheobservedandsimulatedmonthlyflowsduringcalibration.
Month Dvcalibration(%)Jan -24.70Feb 33.57Mar 22.24Apr 11.29May -4.53Jun -11.92Jul 58.63Aug 89.64Sep 80.43Oct 151.13Nov 56.03Dec -9.18
29
Figure2.18.Comparisonoftheobservedandsimulatedmeanmonthlyflowduringcalibrationat
Hoshangabad.
Based on the visualisation of Figures 2.17 and 2.18 aswell as the percentage difference in
volumesgiveninTable2.10,itisveryclearthatthemodelisoverestimatingtheflowsduring
themonthsofthemonsoonseason.Thereasonforthiscanbeattributedtothefactthatthe
modelhasbeenruninthenormalmodewithoutincorporatingthereservoircharacteristics.
As the three major dams store large volumes of water, the water ultimately reaching the
gaugingsiteatHoshangabadonlycomprisesofthedamreleasesintotheriverandtheflows
fromtheintermediatesub-catchments.Assuch,theobservedflowsarealwayslessthanthe
simulatedflowswhichconsidersthebasintobevirginwithouttheimpactofanydams.The
samepatternisobservedduringthevalidationperiod2007–2010asdepictedbyFigure2.19,
which gives the comparison of the daily observed and simulated flows during validation;
Figure2.20,whichgivesthecomparisonofthemonthlyobservedandsimulatedflowsduring
validation; Table 2.11,which gives the percentage difference in volumes during validation;
andFigure2.21,whichgivesthemeanmonthlyflowduringvalidation.
Themodelresultsareencouragingconsideringtheminimaldatainputs.Furthermore,thereis
ample scope for substantial improvements in the simulation of stream flow in the basin,
through the incorporation of reservoir characteristics and adopting the multi-site auto-
calibrationapproach.
30
Figure 2.19. Comparison of the observed and simulated daily flow during validation at
Hoshangabad.
Figure 2.20. Comparison of the observed and simulated monthly flow during validation at
Hoshangabad.
31
Table2.11.Differenceinvolumeintheobservedandsimulatedmonthlyflowsduringcalibration.
Month Dvvalidation(%)Jan 20.88Feb 8.43Mar 7.00Apr -31.63May -42.73Jun -48.81Jul 32.94Aug 185.73Sep 93.97Oct 262.93Nov 135.05Dec 85.14
Figure2.21.Comparisonoftheobservedandsimulatedmeanmonthlyflowduringvalidationat
Hoshangabad.
2.3. HydrologicalmodellingoftheUpperNarmada:Summaryandongoingwork
AMIKE SHEmodel of the UpperNarmada (up to theHoshangabad gauging site) has been
developed that simulatesobserved flows reasonablywell, followingamulti-site calibration.
Likewise, the GWAVA model has been applied over the same area, with the minimal
compulsory inputs of physical parameters, general water demands and climate data. The
manualcalibrationapproachhasbeenusedbyvaryingoneparameteratatimeandobtaining
32
abestfitofthesimulatedflowswiththeobservedflows.InthecaseoftheMIKESHEmodel,
ongoingworkwillimprovetherepresentationofirrigationwithinthemodel,throughgaining
moredetailedinformationonirrigationwithinthebasinfromCEH.Informationpertainingto
theirrigatedcropsinthecommandareaswillalsobeincorporatedwithintheGWAVAmodel.
Inbothmodels,theinclusionofreservoirpropertiespertainingtothethreemajordamswill
be investigated inorder toprovidemore realistic simulationofwater resourceswithin the
basin.Thiswill also give an insight into the change in the flow regime, if any, due to these
interventions in theriverbasin.Further improvements to theGWAVAmodelwillbesought
throughuseoftheauto-calibrationapproach.Multi-sitecalibrationofthemodelwillalsobe
undertakenforthebettersimulationofflowsatalltheavailablegaugingsites.Aftergaining
confidenceinboththeMIKESHEandtheGWAVAmodelsbasedonthevariousperformance
criteria, these models shall be used thereafter to assess the potential impacts of climate
changeonthefuturewaterresourcesoftheNarmadabasin,usinganinter-modelcomparison
approach.Dependingon theavailabilityof resources, themodellingexercise canbe further
extended to the whole of Narmada basin, which, however shall be a challenging task,
consideringthelargenumberofexistingandproposedlargedamsinthebasin.
3. GangaMetagenomePilotStudy
TheGangaformsinUttarakhandastheconfluenceofAlaknandaandBhagirati,twomountain
streamsthatoriginateatHimalayanglaciers.Ittravelsalengthof2525kmthroughtheNorth
Indian plain to the Bay of Bengal (Jain et al., 2007) serving major urban centres such as
Haridwar, Kanpur and Varanasi on its course. This metagenomic pilot study covers 13
sampling points (Figure 3.1) on the first 120 km of the Ganga to capture four seasonal
snapshots of the river ecology upstream and downstream of expected pollution hotspots.
MethodsandpreliminaryresultsaredescribedinSections3.1and3.2.
3.1. Methods
3.1.1. Sampling
Samplingpoints(Figure3.1,Table3.1)werelocatedup-anddownstreamoflikelypollution
hotspotssuchasvillagesandtowns,andwheretributariesjointhemainriverchannel.
33
Figure3.1.SamplingpointsontheGanga.
Watercolumnsamplesweretakenatfourpointsintime:beforetheMonsooninAprilwhen
theriver’swater levelwasat its lowest,aftertheMonsooninOctoberwhenthewater level
wasatitshighestandbetweenthosetwopointsinJuneandDecember.InAprilandOctober,
all 13 siteswere sampled over three days. InDecember and June, a subset of 5 siteswere
sampledononedayfortechnicalreasons. Paralleltotakingwatercolumnsamples inApril
andOctober,thesedimentwassampledatall13sites.
Toperformthewatercolumnsampling,a low-costsetupwasused,comprisingapressure-
resistantwaterbottle,pressurizedwiththehelpofacyclepump,0.22μsterivexfiltersandthe
preservativeRNAlater(ThermoFisherScientific,Loughborough,UKorAMBIONInc.,Austin,
Texas),asdescribedinLehmann(2016).Thefilterwasputinasterilesamplingbag(Thermo
FisherScientific,Loughborough,UK)andstoredinacoolbox.
34
Table3.1.GangaRiversamplingpointlocations.
River/SamplingPoint Lat/Long FurtherinformationRiverBhagirathi/DevPrayag 30.147319,
78.597530aboveconfluence
RiverAlaknanda/DevPrayag 30.15222,78.60012
aboveconfluence
RiverGanga/Bageshwar 30.11375,78.58815
belowconfluence
RiverHanvel/Shivpuri 30.133964,78.389662
Gangatributary
RiverGanga/Shivpuri 30.131812,78.390954
belowHanvel/aboveRishikesh
RiverGanga/TriveniGhat/Rishikesh
30.102782,78.299777
belowRishikesh
RiverGanga/ShantiKunjGhat,Bhupatwala
29.963564,78.180650
aboveHaridwar
RiverGanga/Bisanpour 29.85628,78.14922
belowHardiwar(leftbranch)
RiverGanga/SheetalKhedra 29.82186,78.18161
belowHaridwar(rightbranch)
RiverGanga/Balawali 29.64122,78.10194
abovebarrage(flowing)
RiverGanga/MadhyaGangaBarrageu/s
29.37458,78.04128
abovebarrage(reservoir)
RiverGanga/MadhyaGangaBarraged/s
29.37227,78.04141
belowbarrage
RiverSolani/Mukeempur 29.780465,77.961902
comparisonriver
Forthesedimentsampling,sedimentwasretrievedusingacleanbucketattachedtoarope.
Thebucketwasthrownintothewateranddraggedalongtheriverfloortocatchandretrieve
surface sediment as described in Clark (2003). The sediment sample, pooled andmixed in
that way, was then scooped out of the bucket with a sterile spatula, placed into a sterile
samplingbag,coveredinRNAlaterandstoredinacoolbox.Onarrival inthelaboratorythe
sampleswerestoredat-20°Cuntilextraction.
3.1.2. DNAextraction
ToextractDNAfromthesedimentsamples,thefollowingwereaddedtothepooledsample:
300μloflysisbuffer(100mMNaCl,500mMTris(pH8),10%(w/v)sodiumdodecylsulfate,
2mgml-1proteinaseK,2mgml-1lysingenzymemix(bothSigma-Genosys,Gillingham,UK))
and300μlofNaH2PO4(pH8.0).TheDNAwasincubatedina55°Cwaterbathfor30minand
mixed every 10 min., before adding 80 μl of prewarmed 10% CTAB solution (65°C),
35
incubating at 65°C for 10 minutes and adding 680μl chloroform:isoamyl alcohol (24:1
vol/vol).Thetubeswerecentrifugedfor5minutesat14000rpm.Theaqueoustoplayerwas
aspiratedintoanewtubeandtheDNAprecipitatedbyaddinga300%(w/v)PEG/NaClmix
(30%(w/v)PEG8000,1MNaCl), leavingthesamplesonthebenchfor1h(afterPaithankar
and Prasad, 1991). The samples were then centrifuged (12 per treatment) for 10 min at
14000rpm.ThesupernatantwasdiscardedandtheDNApelletswerewashedbyadding300
μl70%chilledethanol.Thetubeswerecentrifugedagain,theethanolwasdiscardedandthe
tubes were left to dry in a laminar flow cabinet until the ethanol had evaporated. 50 μl
ultrapurewaterwasaddedandtheDNAwaslefttoresuspendfor1honthebench.Forthe
watercolumnsamples,theMoBioPowerwaterDNAextractionkit(MoBio,Carlsbad,US)was
used, followingtheprotocolbutaddingthesamePEGprecipitationstepasdescribedabove
afteritem14oftheextractionprotocol,tocleantheDNAfromRNAlaterresidues.
3.1.3. Sequencing
Afterextraction,thesamplesweresenttoLGCGenomics(Berlin,Germany)formetagenomic
shotgunand16SsequencingontheIlluminaNextSeqplatform.
3.2. Preliminaryresults
The field sampling methods were successfully tested and all samples yielded sufficient
amounts of high-quality DNA. The DNA is currently being sequenced. The DNA library
preparationwaswithoutcomplication.
3.3. Summaryandongoingwork
Samplinghasbeenundertakenonthefirst100kmoftheUpperGangabasintoinvestigatethe
impactofthefirstmajorurbancentresandengineeringprojectsontheecologyandmicrobial
functioningoftheGanga.TheeDNAsamplingcampaign,carriedoutusinglow-costmethods
suitable to theunfavourable conditionsencountered in Indiadue tohigh temperaturesand
remotelocation,yieldedgoodresults.Thedataresultingfromthesequencingthatiscurrently
underway will be pre-processed and analysed on the NERC EOS cluster with the Qiime
(qiime.org) and Kraken (https://ccb.jhu.edu/software/kraken/) pipelines, followed by
analysis tools such as MaAsLin (https://huttenhower.sph.harvard.edu/maaslin), LEfSe
36
(https://bitbucket.org/biobakery/wiki/lefse) and MEGAN (http://ab.inf.uni-tuebingen.de
/software/megan/)fortaxonomicandfunctionalprofiling.
4. Conclusions
Thisprojecthassuccessfullydeveloped twomodelsof theNarmadaBasinusingalternative
hydrologicalmodelcodes.Performanceofthesemodelsisgenerallygoodalthoughon-going
work will seek to fine tune performance through the incorporation of dams (although as
noted above details of how existing dams operate are not yet available). The models are
providing the foundations for subsequent work, already underway with additional NERC
funding,whichfocusesonassessingtheimpactsofalternativefutureclimatechangescenarios
usingtheCMIP5GCMensemble.Modelresultswillbeassessedusingtheenvironmentalflow
approach described by Thompson et al. (2014b). Thiswork also expands the geographical
focustoincludetheBarak-KushiyaraRiverBasinthatspanstheIndian/Bangladeshiborder
(using a modified SWAT model – Rahman et al., in press). Both the Namada and Barak-
Kushiyara(aswellastheMekongandUpperNiger)featureinaproposal inpreparationfor
thecurrentNERCcall“UnderstandingthePathwaystoandImpactsofa1.5°CRiseinGlobal
Temperature”.Thisproposedprojectwilluseexistingmodelstocomparefutureprojections
of river flowandenvironmental flow characteristics for climate change scenarios involving
1.5°Cand2.0°Cchangesinglobalmeantemperature.
TheeDNAsamplingcampaignfortheUpperGangaisthefirstof itskindandhasdeveloped
approachesthataresuitabletotheIndiancontext.Metagenomicshotgunand16Ssequencing
arecurrentlyunderwayandresultswillinformtheuseoftheseapproachesforbiomonitoring
ofmicro-andmacrobiotaand investigatewaterquality issueswithin this, andother Indian
riversystems.
5. References
Acreman,M. and Dunbar,M.J. (2004) 'Defining environmental river flow requirements – areview',HydrologyandEarthSystemSciences,8,861–876.
Ahrends,H.,Mast,M.,Rodgers,C.andKunstmann,H.(2008)'Coupledhydrological–economicmodellingforoptimisedirrigatedcultivationinasemi-aridcatchmentofWestAfrica',EnvironmentalModelling&Software,23,385–395.
Allen,R.G.,Pereira,L.S.,Raes,D.andSmith,M.(1998)Cropevapotranspiration-Guidelinesforcomputingcropwaterrequirements-FAOIrrigationanddrainagepaper56,Rome:FAO.
37
Andersen,J.,Refsgaard,J.C.andJensen,K.H.(2001)'DistributedhydrologicalmodellingoftheSenegal River Basin –model construction and validation', Journal ofHydrology,247,200–214.
Chow,V.T.(1959)OpenChannelHydraulics,NewYork:McGraw-Hill.
Clapp, R.B. and Hornberger, G.M. (1978) 'Empirical equations for some soil hydraulicproperties',WaterResourcesResearch,14,601–604.
Clark,M.J.R.(2003)BritishColumbiaFieldSamplingManual.Victoria,BC,Canada:Water,Air,andClimateChangeBranch,MinistryofWater,Land,andAirProtection.
DHI-WE(2009)MIKESHEUserManual.Volume2:ReferenceGuide,Hørsholm:DHIWaterandEnvironment.
Frei, C., Christensen, J.H., Déqué, M., Jacob, D., Jones, R.G. and Vidale, P.L. (2003) 'Dailyprecipitationstatisticsinregionalclimatemodels:EvaluationandintercomparisonfortheEuropeanAlps',JournalofGeophysicalResearch,108(D3),4124.
Frei,C.andSchär,C.(1998)'AprecipitationclimatologyoftheAlpsfromhigh-resolutionrain-gaugeobservations',InternationalJournalofClimatology,18,873–900.
Gosling, S.N., Taylor, R.G., Arnell, N.W. and Todd, M.C. (2011) 'A comparative analysis ofprojected impactsof climatechangeonriver runoff fromglobalandcatchment-scalehydrologicalmodels',HydrologyandEarthSystemSciences,15,279–294.
Government of India Ministry of Water Resources (2014)Narmada Basin [Online] WaterResources Information System of India Available from: http://www.india-wris.nrsc.gov.in/wris.html[Accessed23February2016].
Graham,D.N.andButts,M.B.(2005)'FlexibleintegratedwatershedmodelingwithMIKESHE',In:Singh,V.P.andFrevert,D.K.(eds.)WatershedModels,BocaRaton:CRCPress,245–272.
Hagemann,S.,Chen,C.,Clark,D.B.,Folwell,S.,Gosling,S.N.,Haddeland,I.,Hanasaki,N.,Heinke,J., Ludwig, F., Voss, F. andWiltshire, A.J. (2013) 'Climate change impact on availablewaterresourcesobtainedusingmultipleglobalclimateandhydrologymodels',EarthSystemDynamics,4,129–144.
Havnø, K., Madsen, M.N. and Dørge, J. (1995) 'MIKE 11 – A generalized river modellingpackage', In: Singh, V.P. (ed.) Computer Models of Watershed Hydrology, HighlandsRanch,Colorado:WaterResourcesPublications,733–782.
Henriksen, H.J., Troldborg, L., Højberg, A.L. and Refsgaard, J.C. (2008) 'Assessment ofexploitablegroundwaterresourcesofDenmarkbyuseofensembleresourceindicatorsandanumerical groundwater–surfacewatermodel', Journal ofHydrology,348,224–240.
Henriksen,H.J., Troldborg, L.,Nyegaard,P., Sonnenborg,T.O.,Refsgaard, J.C. andMadsen,B.(2003) 'Methodology for construction, calibration and validation of a nationalhydrologicalmodelforDenmark',JournalofHydrology,280,52–71.
38
Ho, J.T., Thompson, J.R. and Brierley, C. (2015) 'Projections of hydrology in the Tocantins-Araguaia Basin, Brazil: uncertainty assessment using the CMIP5 ensemble',HydrologicalSciencesJournal,AuthorManuscriptOnline.
Immerzeel, W.W., Beek, L.P.H., Konz, M., Shrestha, A.B. and Bierkens, M.F.P. (2012a)'HydrologicalresponsetoclimatechangeinaglacierizedcatchmentintheHimalayas',ClimaticChange,110,721–736.
Immerzeel,W.W.,Pellicciotti,F.andShrestha,A.B.(2012b)'GlaciersasaProxytoQuantifytheSpatial Distribution of Precipitation in the Hunza Basin', Mountain Research andDevelopment,32,30–38.
India-WRIS (2013a) Bargi (Rani Avanti Bai Lodhi Sagar) Major Irrigation Project JI00740[Online] Available from: http://india-wris.nrsc.gov.in/wrpinfo/index.php?title=Bargi(Rani_Avanti_Bai_Lodhi_Sagar)Major_Irrigation_Project_JI00740 [Accessed23February2016].
India-WRIS (2013b) Barna Major Irrigation Project JI00745 [Online] Available from:http://india-wris.nrsc.gov.in/wrpinfo/index.php?title=Barna_Major_Irrigation_Project_JI00745[Accessed23February2016].
India-WRIS (2015) Narmada [Online] Available from: http://india-wris.nrsc.gov.in/wrpinfo/index.php?title=Narmada[Accessed23February2016].
Jain,S.K.,Storm,B.,Bathurst,J.C.,Refsgaard,J.C.andSingh,R.D.(1992)'ApplicationoftheSHEtocatchmentsinIndiaPart2.FieldexperimentsandsimulationstudieswiththeSHEontheKolarsubcatchmentoftheNarmadaRiver',JournalofHydrology,140,25–47.
Jain, S.K., Agarwal, P.K. and Singh, V.P. (2007) 'Ganga Basin', In: Hydrology and WaterResourcesofIndia,Dordrecht:SpringerNetherlands,333–418.
Johnston,R.andSmakhtin,V.(2014)'HydrologicalModelingofLargeRiverBasins:HowMuchisEnough?',WaterResourcesManagement,28,2695–2730.
Kalantari, Z., Lyon, S.W., Folkeson, L., French, H.K., Stolte, J., Jansson, P.E. and Sassner, M.(2014)'Quantifyingthehydrologicalimpactofsimulatedchangesinlanduseonpeakdischarge in a small catchment',The Science of the total environment,466-467, 741–754.
Lehmann, K. (2016) ‘Sampling of riverine or marine bacterial communities in remotelocations: from field to publication’, In: Bourlat, S.J., (ed.)Marine genomics: methodsand protocols. New York, Springer Science and Business Media, 1–18. (Methods inMolecularBiology).(July2016).
Li,Y.,Thompson,J.R.,Li,H.(2016)‘Impactsofspatialclimaticrepresentationonhydrologicalmodel calibration and prediction uncertainty: A mountainous catchment of ThreeGorgesReservoirRegion,China’,Water,8,73.
Loucks,D.P.andvanBeek,E.(2005)WaterResourcesSystemsPlanningandManagement:AnIntroduction to Methods, Models and Applications, Paris: United Nations Educational,ScientificandCulturalOrganization.
39
Masood,M.,Yeh,P.J.F.,Hanasaki,N. andTakeuchi,K. (2015) 'Model studyof the impactsoffuture climate change on the hydrology of Ganges–Brahmaputra–Meghna basin',HydrologyandEarthSystemSciences,19,747–770.
McCartney,M.P.andAcreman,M.C.(2009)'WetlandsandWaterResources',In:Maltby,E.andBarker,T.(eds.)TheWetlandsHandbook,Chichester:Wiley-Blackwell.
Moore,R.J. (1985) ‘Theprobability-distributedprincipleandrunoffproductionatpointandbasinscales’,HydrologicalSciencesJournal,30,273–297.
Moore,R.J.(2007)'ThePDMrainfall-runoffmodel',HydrologyandEarthSystemSciences,11,483–499.
Nash,I.E.andSutcliffe,I.V.(1970)'Riverflowforecastingthroughconceptualmodels',JournalofHydrology,10,282–290.
Norman,D.W.andDixon,J.(1995)Sustainabledrylandcroppinginrelationtosoilproductivity-FAOsoilsbulletin72,Rome:FoodandAgricultureOrganizationoftheUnitedNations.
Pai, D., Sridhar, L., Rajeevan, M., Sreejith, O.P., Satbhai, N.S. and Mukhopadhyay, B. (2014)'Developmentofanewhighspatialresolution(0.25°×0.25°)longperiod(1901-2010)daily gridded rainfall data set over India and its comparisonwith existing data setsovertheregion',Mausam,65,1–18.
Paithankar,K.R. andPrasad,K.S.N. (1991) 'PrecipitationofDNAbypolyethyleneglycol andethanol',NucleicAcidsResearch,19,1346.
Patel, J., Patel, H. and Bhatt, C. (2014) 'ECALTOOL: fuzzy logic based computer program tocalibrate the Hargreaves equation for accurate estimation of evapotranspiration',AgriculturalEngineeringInternational:CIGRJournal,16,245–250.
Payasi, Y.K. (2015) 'Diversity andDistribution of Phytoplankton in the RiverHiran',GlobalJournalforResearchAnalysis,4.
Prudhomme, C. and Davies, H. (2009) 'Assessing uncertainties in climate change impactanalysesontheriverflowregimesintheUK.Part1:baselineclimate',ClimaticChange,93,177–195.
Rahman,M.M.,Thompson,J.R.,Flower,R.J.(Inpress).AnenhancedSWATwetlandmoduletoquantify hydraulic interactions between riparian depressional wetlands, rivers andaquifers.EnvironmentalModellingandSoftware.
Rajaguru,S.N.,Gupta,A.,Kale,V.S.,Mishra,S.,Ganjoo,R.K.,Ely,L.L.,Enzel,Y.andBaker,V.R.(1995) 'Channel form and processes of the flood-dominated Narmada River, India',EarthSurfaceProcessesandLandforms,20,407–421.
Räsänen, T.A., Koponen, J., Lauri, H. and Kummu, M. (2012) 'Downstream HydrologicalImpacts of Hydropower Development in the UpperMekong Basin',Water ResourcesManagement,26,3495–3513.
Refsgaard, J.C. and Henriksen, H.J. (2004) 'Modelling guidelines––terminology and guidingprinciples',AdvancesinWaterResources,27,71–82.
40
Refsgaard, J.C., Storm, B. and Clausen, T. (2010) 'Système Hydrologique Europeén (SHE):review and perspectives after 30 years development in distributed physically-basedhydrologicalmodelling',HydrologyResearch,41,355–377.
Richter,B.D.,Baumgartner,J.V.,Wigington,R.andBraun,D.P.(1997)'Howmuchwaterdoesariverneed?',FreshwaterBiology,37,231–249.
Singh, C.R., Thompson, J.R., Kingston, D.G. and French, J.R. (2011) 'Modelling water-leveloptions for ecosystem services and assessment of climate change: Loktak Lake,northeastIndia',HydrologicalSciencesJournal,56,1518–1542.
Srivastava,A.K.,Rajeevan,M.andKshirsagar, S.R. (2009) 'Developmentofahigh resolutiondaily gridded temperature data set (1969-2005) for the Indian region',AtmosphericScienceLetters,10,249–254.
Stisen, S., Jensen, K.H., Sandholt, I. and Grimes, D.I.F. (2008) 'A remote sensing drivendistributedhydrologicalmodelof theSenegalRiverbasin', JournalofHydrology,354,131–148.
Tan, B., Ng, C., Nshimyimana, J.P., Loh, L.L., Gin, K.Y.H. and Thompson, J.R. (2015) ‘Next-generation sequencing (NGS) for assessment of microbial water quality: currentprogress, challenges, and future opportunities’, Frontiers in Microbiology, 6: 1027,doi:10.3389/fmicb.2015.01027.
Thompson, J.R., Green, A.J. andKingston,D.G. (2014a) 'Potential evapotranspiration-relateduncertainty in climate change impacts on river flow: An assessment for theMekongRiverbasin',JournalofHydrology,510,259–279.
Thompson,J.R.,Green,A.J.,Kingston,D.G.andGosling,S.N.(2013)'Assessmentofuncertaintyinriver flowprojections for theMekongRiverusingmultipleGCMsandhydrologicalmodels',JournalofHydrology,486,1–30.
Thompson, J.R., Laizé, C.L.R., Green, A.J., Acreman,M.C. andKingston,D.G. (2014b) 'Climatechange uncertainty in environmental flows for the Mekong River', HydrologicalSciencesJournal,59,935–954.
Thomsen,P.F.,Kielgast,J.,Iversen,L.L.,Wiuf,C.,Rasmussen,M.,Gilbert,M.T.P.,Orlando,L.andWillerslev, E. (2012) ‘Monitoring endangered freshwater biodiversity usingenvironmentalDNA’,MolecularEcology,21,2565–2573.
Vázquez, R.F., Feyen, L., Feyen, J. and Refsgaard, J.C. (2002) 'Effect of grid size on effectiveparametersandmodelperformanceoftheMIKE-SHEcode',HydrologicalProcesses,16,355–372.
Velázquez,J.A.,Schmid,J.,Ricard,S.,Muerth,M.J.,GauvinSt-Denis,B.,Minville,M.,Chaumont,D., Caya, D., Ludwig, R. and Turcotte, R. (2013) 'An ensemble approach to assesshydrological models' contribution to uncertainties in the analysis of climate changeimpactonwaterresources',HydrologyandEarthSystemSciences,17,565–578.
Vieux,B.E.(2004)DistributedHydrologicModelingUsingGIS,Dordrecht:KluwerAcademic.
41
Wijesekara, G.N., Farjad, B., Gupta, A., Qiao, Y., Delaney, P. and Marceau, D.J. (2014) 'Acomprehensiveland-use/hydrologicalmodelingsystemforscenariosimulationsintheElbowRiverwatershed,Alberta,Canada',Environmentalmanagement,53,357–381.
Wijesekara, G.N., Gupta, A., Valeo, C., Hasbani, J.G., Qiao, Y., Delaney, P. and Marceau, D.J.(2012) 'Assessingtheimpactoffutureland-usechangesonhydrologicalprocessesinthe Elbow Riverwatershed in southern Alberta, Canada', Journal of Hydrology,412-413,220–232.
Winchell, M., Srinivasan, R., Di Luzio, M. and Arnold, J. (2013) ArcSWAT Interface forSWAT2012:User'sGuide,Temple,Texas:BlacklandResearch&ExtensionCenterTexasAgriLifeResearch;Grassland,SoilandWaterResearchLaboratoryUSDAAgriculturalResearchService.
top related