Citation
El Niño, Climate, and Cholera Associations in Piura, Peru, 1991–2001: A Wavelet Analysis

Material Information

Title:
El Niño, Climate, and Cholera Associations in Piura, Peru, 1991–2001: A Wavelet Analysis
Series Title:
EcoHealth
Creator:
Ramírez, Ivan J.
Grady, Sue C.
Place of Publication:
US
Publisher:
Springer
Publication Date:

Subjects

Genre:
Journal Article ( sobekcm )

Notes

Abstract:
In Peru, it was hypothesized that epidemic cholera in 1991 was linked to El Niño, the warm phase of El Niño–Southern Oscillation. While previous studies demonstrated an association in 1997–1998, using cross-sectional data, they did not assess the consistency of this relationship across the decade. Thus, how strong or variable an El Niño–cholera relationship was in Peru or whether El Niño triggered epidemic cholera early in the decade remains unknown. In this study, wavelet and mediation analyses were used to characterize temporal patterns among El Niño, local climate variables (rainfall, river discharge, and air temperature), and cholera incidence in Piura, Peru from 1991 to 2001 and to estimate the mediating effects of local climate on El Niño–cholera relationships. The study hypothesis is that El Niño-related connections with cholera in Piura were transient and interconnected via local climate pathways. Overall, our findings provide evidence that a strong El Niño–cholera link, mediated by local hydrology, existed in the latter part of the 1990s but found no evidence of an El Niño association in the earlier part of the decade, suggesting that El Niño may not have precipitated cholera emergence in Piura. Further examinations of cholera epicenters in Peru are recommended to support these results in Piura. For public health planning, the results may improve existing efforts that utilize El Niño monitoring for preparedness during future climate-related extremes in the region.
Acquisition:
Collected for Auraria Institutional Repository by the Self-Submittal tool. Submitted by Ivan J. Ramírez.
Publication Status:
Published

Record Information

Source Institution:
Auraria Institutional Repository
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

ElNin ˜ o,Climate,andCholeraAssociationsinPiura,Peru, 1991–2001:AWaveletAnalysis Iva ´ nJ.Ram ´ rez 1,2 andSueC.Grady 3 1 InterdisciplinaryScienceProgram,TheNewSchool,65W11thStreet,NewYork,NY10011 2 TishmanEnvironmentandDesignCenter,TheNewSchool,NewYork,NY 3 DepartmentofGeography,MichiganStateUniversity,EastLansing,MI Abstract: InPeru,itwashypothesizedthatepidemiccholerain1991waslinkedtoElNin ˜ o,thewarmphaseof ElNin ˜ o–SouthernOscillation.Whilepreviousstudiesdemonstratedanassociationin1997–1998,usingcrosssectionaldata,theydidnotassesstheconsistencyofthisrelationshipacrossthedecade.Thus,howstrongor variableanElNin ˜ o–cholerarelationshipwasinPeruorwhetherElNin ˜ otriggeredepidemiccholeraearlyinthe decaderemainsunknown.Inthisstudy,waveletandmediationanalyseswereusedtocharacterizetemporal patternsamongElNin ˜ o,localclimatevariables(rainfall,riverdischarge,andairtemperature),andcholera incidenceinPiura,Perufrom1991to2001andtoestimatethemediatingeffectsoflocalclimateonElNin ˜ o– cholerarelationships.ThestudyhypothesisisthatElNin ˜ o-relatedconnectionswithcholerainPiurawere transientandinterconnectedvialocalclimatepathways.Overall,ourndingsprovideevidencethatastrongEl Nin ˜ o–choleralink,mediatedbylocalhydrology,existedinthelatterpartofthe1990sbutfoundnoevidenceof anElNin ˜ oassociationintheearlierpartofthedecade,suggestingthatElNin ˜ omaynothaveprecipitated choleraemergenceinPiura.FurtherexaminationsofcholeraepicentersinPeruarerecommendedtosupport theseresultsinPiura.Forpublichealthplanning,theresultsmayimproveexistingeffortsthatutilizeElNin ˜ o monitoringforpreparednessduringfutureclimate-relatedextremesintheregion. Keywords: ElNin ˜ o,ElNin ˜ o–SouthernOscillation,cholera,climate,wavelet,mediation I NTRODUCTION Since1991,epidemiccholerahasemergedtwiceinthe westernhemisphere,contributingtoapproximately2millioncasesintheregion(PanAmericanHealthOrganization 2008 , 2014 ).TherstemergencebeganinPeruin1991, spreadtoSouthandCentralAmerica,andlastedapproximatelyadecade.Thesecondemergence,whichisongoing, spreadfromHaitiin2010totheDominicanRepublic, Cuba,andMexico(Mooreetal. 2014 ).InPeru,similarto inHaiti(Jutlaetal. 2013 ),studiessuggestedthatcholera emergencewastriggeredbyclimateimpacts.InPeru,itwas hypothesizedthatepidemiccholerawaslinkedtoElNin ˜ o, thewarmphaseofElNin ˜ o–SouthernOscillation(ENSO) (Epsteinetal. 1993 ;Colwell 1996 ).ENSOisaclimaticcycle intheequatorialPacicOceanthataffectsglobaltolocal weatherpatternsevery2–7years. Todate,moststudiesexaminingthishypothesishave focusedonthe1997–1998ElNin ˜ o,ndingpositivecorrelationsbetweenlocalclimateandcholeraincidence.For Publishedonline:January29,2016 Correspondenceto: Iva ´ nJ.Ram ´ rez,e-mail:ramirezi@newschool.edu EcoHealth13,83–99,2016 DOI:10.1007/s10393-015-1095-3 OriginalContribution 2016InternationalAssociationforEcologyandHealth

PAGE 2

example,Speelmonetal.( 2000 )andGiletal.( 2004 ) showedthatelevatedcoastalairandseatemperatureswere associatedwithcholeraepidemicsinLima,thecapitalof Peru.Morerecently,Ramirez( 2015 )demonstratedthat rainfall,inadditiontoairandseatemperatures,increased choleraatadistrictlevelinPiura,Peru(900kmnorthof Lima).Whilethesestudiesdemonstratedanassociation withElNin ˜ ousingcross-sectionaldata,i.e.,atonepointin time,theydidnotassesstheconsistencyofthisrelationship acrossthedecade.Furthermore,previousstudieshavenot investigatedtheimpactofElNin ˜ ooncholeraintheearly 1990s.Understandingtheseyearsisimportantbecauseit representsaperiodwhencholeraemerged,andreportedly, whenElNin ˜ o’stimingandeffectswereunclear(Ramirez etal. 2013 ).Thus,howstrongorvariableanElNin ˜ o– cholerarelationshipwasinPeruorwhetherElNin ˜ otriggeredtheepidemicin1991remainsunknown.Suchtemporalinformationmaybecriticalforpublichealth planningintheregion,whereElNin ˜ omonitoringis important. Theobjectiveofthisstudywas,therefore,toestimate thetemporalrelationshipsamongElNin ˜ o,climate,and choleraincidencefrom1991to2001.Wefocusedonthe northernregionofPiura,Peruwhereindexcaseswerereportedin1991(Riesetal. 1992 ).ToexamineElNin ˜ o– choleraassociations,weemployedwaveletanalysisto characterizebivariatetemporalpatternsbetweenclimate variablesandcholeracases.Waveletanalysisisamethod usedingeophysicalstudiestoexaminetheoscillating componentsandfrequencyofrelationshipsbetweenocean andatmosphericprocesses,includingENSO(Torrenceand Compo 1998 ).Inparticular,itisusefultostudycomplex relationships,whosetemporalpropertiesexhibitnon-stationarity,meaningtheirperiodiccomponentsandassociationschangeovertime(Cazellesetal. 2007 ).Wavelet analysisprovidesanadvantageovertraditionaltime-series approaches,suchasregressionandFourieranalyses,which assumeconsistentassociations(Cazellesetal. 2007 ).Nonstationarityinclimate–diseaserelationshipshasbeenreportedinseveralglobalstudies,includingcholerainGhana (ConstantindeMagnyetal. 2006 )andBangladesh(Hashizumeetal. 2013 ),dengueinThailand(Cazellesetal. 2005 ),andLeishmaniasisinCostaRica(ChavesandPascual 2006 ).Thisstudywilldifferfrompreviousstudiesby notonlyusingawavelettime-seriesapproachbutalso combiningwaveletwithmediationanalysistoestimate impactsoflocalclimatevariablesonElNin ˜ o–cholera associationsinPiura.Mediationanalysisisamethodemployedinpsychologyandpublichealthtoestimatethe effectsofinterveningvariablesthroughwhichacausal factoraffectsahealthoutcome(Frazieretal. 2004 ;Grady andRamirez 2008 ).AsstudiesinPeru(Giletal. 2004 ; Ramirez 2015 )andBangladesh(Pascualetal. 2000 , 2002 ; Cashetal. 2008 )suggest,ElNin ˜ o’seffectoncholerais interconnectedvialocalclimateinuences,knownasteleconnections(Glantz 1991 ).Inthisresearch,ouroverarchinghypothesisisthatapotentialElNin ˜ o–choleralinkin Piurawastransientandmediatedbylocalclimatevariables, specicallyrainfall,riverdischarge,andairtemperature.METHODSStudyAreaThestudyareaisPiura,locatedintheDepartmentofPiura innorthwesternPeru.PiuraisahealthadministrativeregionthatbordersthePacicOceanandthefoothillsofthe Andes(Fig. 1 ).Piura’sclimatevariesfromsemi-aridonthe low-lyingcoasttosubtropicalconditionsinthemountainouseast(PAEN/GTZ 2003 ).Figure 2 showsairtemperatureandrainfallattheMiraoresmeteorological station,locatedinthecapitalcityofPiura.Theannual averageairtemperatureis24 C,andmedianrainfallis 43mm,1exceptduringElNin ˜ os,whenrecordestimatesare reported(e.g.,1849mm—1998)(Takahashi 2004 ). Approximately68.0%ofthepopulationlivedonthecoast duringthestudyperiod(InstituteofNationalStatisticsand Information[INEI] 2000 ).DataWeeklycholeradataincludingprobableandconrmed casesfrom1991to2001wereobtainedfromtheDepartmentsofEpidemiologyattheMinistriesofHealth(MINSA)inLimaandPiura,Peru.Thecasedenitionfor probablecholerawas‘‘apersonofanyagewhodeveloped acutewaterydiarrhea,withorwithoutvomiting,with severedehydrationorshockordiedfromacutewatery diarrhea’’(MINSA 2005 ).Aconrmedcaseofcholerawas denedasalaboratoryisolateof Vibriocholerae01 froma patient’sbloodspecimen.Ifaprobablecasewasconrmed within2weeksfollowingtheinitialdiagnosisthecasewas changedfromprobabletoconrmed.Inthisstudy,weekly 1EstimatebasedonMiraoresdatafrom1971–2003withoutElNin ˜ oyears.84I.J.Ram ´ rez,S.C.Grady

PAGE 3

choleraprobableandconrmedcaseswereaggregatedby monthoverthestudytimeperiod( n =132months)and square-roottransformedtodampenpositiveextremevaluespriortowaveletanalysis(Cazellesetal. 2005 ). Monthlyseasurfacetemperature(SST)anomalydata representingENSOintheNin ˜ oRegions3.4and1+2were obtainedfromtheNationalOceanicandAtmospheric Administration(NOAA 2015 )( http://www.cpc.ncep.noaa. gov/data/indices/ ).TheseSSTregionswereselectedbecause ofproximitytothePeruviancoast(Nin ˜ o1+2)andareasof convection(Nin ˜ o3.4)intheequatorialPacicOcean.To interpretourresults,weidentiedElNin ˜ oepisodesbased onthecommonlyusedONI-indexassociatedwithNin ˜ o3.4 ( http://www.cpc.ncep.noaa.gov/products/analysis_monitor ing/ensostuff/ensoyears.shtml ).Localclimatedatacollected includedmonthlyaveragePaitaSSTanomaly,airtemperatureanomaly( Tmean),rainfall,andriverdischarge(proxy forriverow).Thesedatacomefromamonitoringstation inthePortofPaita,andtheMiraoresstation,mentioned earlier.Allclimatedatawerecollectedbeginningin1971in ordertocapturetemporaltrendspriortothe1991cholera epidemic.Forafurtherdescriptionofthesedata,seeRamirez( 2015 ).Rainfallandriverdischargedatawerealso square-roottransformedpriortoanalysis. Figure1. MapofPiurahealthadministrationregioninPeru. ElNin ˜ oandCholeraAssociationsinPiura,Peru85

PAGE 4

WaveletAnalysisTocharacterizeElNin ˜ o,localclimate,andcholerarelationships,threetypesofwaveletanalyseswereimplemented.First,eachtimeserieswascharacterizedusing continuouswavelettransform(CWT)todecomposeeach seriesintotime–frequencyspaceandidentifyareasofhigh/ lowperiodicity,i.e.,variabilityandfrequency,accordingto period(scalebyyear[s])andtimeinterval(s)(Torrenceand Compo 1998 ).Second,tounderstandtherelationshipbetweentwotimeseries,cross-wavelettransform(XWT)and waveletcoherence(WTC)wereemployed.TheXWTestimatedcommonareaswheretwoseriesco-variedand sharedhigh/lowperiodicity,whiletheWTCmeasuredthe strengthofcovarianceandidentiedareasoflinearcorrelation.BothXWTandWTCanalysesestimateddirectionof relationships(e.g.,in-phaseorout-of-phase)andtemporal lags(indicatedbyarrowsinthewaveletgures).Toaddress serialcorrelation,autoregressionswerecontrolledforusing arst-orderautoregressiveterm,aprocesscommonlyused tomodelgeophysicaltimeseries(TorrenceandCompo 1998 ).WaveletanalyseswereperformedinMatlabR2013a usingscriptswrittenbyGrinstedetal.( 2004 , 2008 )andNg andKwok( 2012 ).MediationAnalysisFollowingthewaveletanalysis,signicanttimeintervalsof coherencewereexploredfurtherformediatingeffectsof localclimateonElNin ˜ o–cholerarelationships.First,temporallags(0–12months)wereexploredtoidentifythebest climatevariable-predictorsofcholera.TheBarronand Kenny’s( 1986 )modeltotestformediatingeffectswas implemented(Fig. 3 ).Therststepwastoshowthatthere wasasignicantrelationshipbetweenSSTandcholera (Pathc).Pathcwasusedasareferencemodelfromwhich subsequentmodelswerecompared.Thesecondstepwasto showthatSSTwassignicantlyassociatedwithlocalclimate(Patha).Thethirdstepwastoshowthatlocalclimate wassignicantlyassociatedwithcholera(Pathb),controllingforSST.Finally,Pathc’measuredthemediating (reduced)effectofSSToncholeraafterlocalclimatewas addedtothemodel.Todetermineifthereducedeffectwas signicant,theAroianversionoftheSobeltestrecommendedbyPreacherandHayes( 2004 )wasimplemented. Mediationanalyseswereconductedusingordinaryleast squares(OLS)regressionmodelsinSPSSv.22(IBM-SPSS 2015 ).RESULTSFrom1991to2001,therewere38,040casesofcholerareportedinPiura,Peru,representing5.2%ofallcasesin Peru.Amajority(83.0%)ofthesecaseswerereportedin therst2yearsoftheepidemic.Oftheremainingcases ( n =6406),amajority(64.7%)werelaterreportedin1998. Figure 4 a–gshowsthetime-series(leftpanel)andCWT analyses(rightpanel)ofclimatevariablesandcholeracases. Waveletanalysisofcholera(Fig. 4 a)revealedmoderateto highperiodicityduringtheonsetoftheepidemicatperiods of1yearandlessfrom1991–1993.Signicantperiodicity, however,wasconnedto1992–1993,duetoedgeeffects (discontinuities)atthemarginsofthewaveletpower spectrum(blackcurvedelimitstheconeofinuence). Thereafter,lowperiodicity(declineincholeracases)was observed,followeddistinctlybymoderateperiodicityin 1997–1998atperiodslessthan1.5years.Althoughcholera periodicityatthisintervalwasnotstatisticallysignicant, Figure2. Monthlyaveragesofairtemperatureandrainfallparametersfor1971–2001attheMiraoresmeteorologicalstationlocated inthecapitalcityofPiura. Figure3. Pathways(a–c,andc’)inthemediationmodeloftheSST andcholerarelationshipadaptedfromBarronandKenny( 1986 ).86I.J.Ram ´ rez,S.C.Grady

PAGE 5

theshapeofthespectracapturedthemajorepidemicreportedin1998.Whycholeraactivityfrom1997to1998was notsignicantwassurprisingandinterestinggivenitwas thethirdlargestepidemicinPiurainthe1990s. IntheSSTanalysis,signicanthighperiodicitywas evidentatperiodsofapproximately3yearsand5years. Thereweresomedifferencestonote,however.Forexample, aquasi-continuouscyclewasobservedintheNin ˜ o3.4region Figure4.a – g Monthlytimeseries( leftpanel )andcontinuouswavelettransformanalyses( rightpanel )ofcholeraandclimatevariables: a choleracases(square-roottransformed); b Nin ˜ o3.4seasurfacetemper ature(SST)anomaly; c Nin ˜ o1+2SSTanomaly; d PaitaSSTanomaly; e airtemperature( Tmean)anomaly; f rainfall(square-roottransformed);and g riverdischarge(square-roottransformed).Thecontinuouswavelet transformisdenotedbyperiod(scaleby year)andacrosstimeintervals.The colorcode forpowervaluesincreasesfrom darkblue (low)to dark red (high),andstatisticalsignicance(95.0%co ndencelevel)isindicatedbyareaswithin thickblackoutlines .The blackcurve delimitsthecone ofinuence(COI),aregioninuencedbyedgeeffects. ElNin ˜ oandCholeraAssociationsinPiura,Peru87

PAGE 6

Figure4. continued88I.J.Ram ´ rez,S.C.Grady

PAGE 7

(Fig. 4 b),particularlyfrom1980to1992(5-yearperiod). WhileintheNin ˜ o1+2regionandcoastalPaita(Fig. 4 c,d), ENSO’sinuencewastransientbutdistinguishedduring signicantElNin ˜ os(e.g.,1972–1973,1982–1983,and1997– 1998).Inaddition,therewasalsosignicanthighperiodicity intheNin ˜ o3.4series(e.g.,1972–1973;1982–1984;and 1992–1999)andtosomedegreeintheNin ˜ o1+2series (1983–1984)atthe1.5-yearperiod.Forairtemperature Figure5.a – f Crosswavelettransform( leftpanel )andwaveletcoherence( rightpanel )analysesbetweenclimat evariablesandcholera: a Nin ˜ o 3.4SSTanomaly; b Nin ˜ o1+2SSTanomaly; c PaitaSSTanomaly;dairtemperature( Tmean)anomaly; e rainfall(square-roottransformed);and f riverdischarge(square-roottransf ormed).Bothanalysesaredenotedbyperiod(scalebyyear)andacrosstimeintervals.Thecolorcodeshows powervalues(crosswavelet)andcoherenceval ues(waveletcoherence)thatincreasefrom darkblue (low)to darkred (high).Thedirection (phase)ofrelationshipsisindicatedbyarrows,assuch: up (climatelags); down (climateleads); right (climate-cholerain-phase);and left (climate-choleraoutofphase).Statis ticalsignicance(95.0%condencelevel)isindicatedbyareaswithinthick blackoutlines .The blackcurve delimitstheconeofinuence(COI),aregioninuencedbyedgeeffects. ElNin ˜ oandCholeraAssociationsinPiura,Peru89

PAGE 8

(Fig. 4 e),areasofsignicanthighperiodicitywerelimitedto twoENSO-relatedintervals:1997–1999and1983–1985at periodsof2–3yearsand4years.Thewaveletanalysesof rainfallandriverdischarge(Fig. 4 f,g)revealedadistinct patternofsignicanthighperiodicitycoincidingwith ENSO-relatedintervals(1972–1973;1982–1984;and1997– 1999),aswellasin1993–1994.Periodicityattheseintervals wasevidentacrossmultipleperiodsatlessthan2years. Figure 5 a–fdisplaystheXWT(leftpanel)andWTC (rightpanel)analysesbetweenclimatevariablesandcholera cases.Overall,theresultsrevealedpatternsassociatedwith twosetsofclimateparameters:(a)Nin ˜ o1+2SST,Paita SST,and Tmean;and(b)Nin ˜ o3.4SST,rainfall,andriver discharge.Set(a)sharedacommonareaofperiodicitywith cholerain1992–1993;whileset(b)sharedcommonareas ofperiodicitywithcholerainthetimeintervals,1992–1993 and1996–1999.Oncethesepatternswereexaminedfurther,theWTCanalysesidentiedsignicantcoherency withcholerainonetimeinterval,1996–1999,withallclimatevariablesatperiodsthatdifferedbyset:set(a)was coherentatintraannualtoannualscalesof.5–1.5years, whereas,set(b)wascoherentatannualtointerannual Figure5. continued90I.J.Ram ´ rez,S.C.Grady

PAGE 9

scalesof1–3years.Thephaserelationshipssuggestpositive associationswhererainfallandriverdischargevariables werecloselysynchronizedwithcholera(in-phase,0 ),while SSTand Tmeanvariablesledcholerabyapproximately45 – 90 ,suggestingatimedelayofseveralmonths. Inaddition,weexaminedSST-localclimateassociationsfrom1971to2001toillustratepotentiallinksina mediationanalysis(seeFig. 10 in Appendix ).Ingeneral, SSThadstrongpositivecoherentassociationswithrainfall, riverdischarge,and Tmean.Theinterannualsignal,namely ENSO,wasclearlyevidentinthecoherencypatterns(e.g., periods > 2.5years).WaveletSummaryIndependently,choleraperiodicitywasdetectedduringthe initialepidemicyears(1992–1993)inPiura.Oncethispatternwasexaminedinrelationtoclimate,coherencywas observedin1996–1999,whichincludestheresurgenceof cholerain1998,atperiodsthatrangedfrom0.5–3years.No signicantclimate–choleracoherencywasobservedinother timeintervals,including1991–1993(choleraemergence). Overall,thestrongestcoherenciesin1996–1999wereassociatedwithrainfallandriverdischargeindicatedbystrong correlationsandconsistentphaserelationshipsacrossmultipleperiods.Ingeneral,rainfallandriverdischargelinks werein-phase(0-to1-monthlag)withcholera,comparedto seaandairtemperatureassociationswithcholerawhichhad longertemporallags( * severalmonths).MediationResultsInformedbythewaveletanalysis,theinterval1996–1999was selectedtoexploremediationfactorsinElNin ˜ o–cholera relationships.Tables 1 , 2 , 3 , 4 ,and 5 showtheOLSstatistical summaries,includingthetemporallagsforeachpredictor. Figures 6 , 7 ,and 8 illustratemediationpathwayresults.The effectsofNin ˜ o3.4SST( b =0.674, P value=0.000),Nin ˜ o 1+2SST( b =0.620, P value=0.000),andPaitaSST ( b =0.673, P value=0.000)oncholeraweresignicantand positive(Pathc).InPatha,allSSTvariablesweresignicantly andpositivelyassociatedwithrainfall,riverdischarge,and Tmean( P value=0.000).Rainfall( b =0.936, P value=0.000),riverdischarge( b =0.894, P value=0.000), and Tmean( b =0.702, P value=0.000)werealsosignicantly associatedwithcholera.Oftheselocalclimateparameters,the strongestmediatorsintheSST–cholerarelationshipswere rainfall [Nin ˜ o3.4( b =0.833, P value=0.000);Nin ˜ o1+2 ( b =0.840, P value=0.0000);andPaita( b =0.809, P value=0.0000)]and riverdischarge [Nin ˜ o3.4( b =0.755, P value=0.000);Nin ˜ o1+2( b =0.818, P value=0.0000);and Paita( b =0.766, P value=0.0000)],controllingSST(Path b). TmeanwasasignicantmediatoronlyintheNin ˜ o3.4SST– cholerarelationship( b =0.453, P value=0.013).TheSobel Testresults(Pathc’)areshowninTable 6 .Thedecreaseinthe effectsofSSTwasstatisticallysignicantandsupportedthe observationsdescribedabove.Ingeneral,theoptimaltemporallagsidentiedintheOLSmodelssupportthetrendsin thewaveletndings:seaandairtemperaturesledcholeraby severalmonths(Nin ˜ o3.4SST—4-monthlag;Nin ˜ o1+2 SST—6-monthlag;PaitaSST—6-monthlag; Tmean— 6-monthlag),whereasrainfallandriverdischargeweresynchronized(rainfall—1-monthlag;andriverdischarge—zeromonthlag). Table1. Pathc:EffectofSST(Lag-Months)onMonthlyCholeraCases,Piura,1996–1999. Predictor(lag) b SE t -ratio P value Nin ˜ o3.4SST(4)0.6740.7786.1890.000 Nin ˜ o1+2SST(6)0.6200.4995.3650.000 PaitaSST(6)0.6730.6166.1650.000 Table2. Patha:EffectofSST(Lag-Months)onLocalClimate Variables,Piura,1996–1999. b SE t -ratio P value Outcome Rainfall Predictor(lag) Nin ˜ o3.4SST(3)0.6090.6095.2090.000 Nin ˜ o1+2SST(4)0.5250.3934.1860.000 PaitaSST(4)0.5700.5044.7010.000 Outcome Riverdischarge Predictor(lag) Nin ˜ o3.4SST(4)0.5701.0294.7010.000 Nin ˜ o1+2SST(6)0.6050.6025.1580.000 PaitaSST(5)0.6180.7795.3320.000 Outcome Airtemperature Predictor(lag) Nin ˜ o3.4SST(0)0.8540.08711.1360.000 Nin ˜ o1+2SST(0)0.9520.03120.9780.000 PaitaSST(0)0.9320.05017.3910.000 ElNin ˜ oandCholeraAssociationsinPiura,Peru91

PAGE 10

DISCUSSIONUsingwaveletanalysis,correlationsbetweencholeraincidenceandElNin ˜ oandlocalclimatewerefoundatmultiple timescales(0.5–3years)inthelatterhalfofthe1990s,but noother-relatedlinkswereobservedatthebeginningofthe epidemicorothertimeintervals.Specically,theanalysis showedthatrainfallandriverdischargewerestrongly associatedwithcholerainPiurain1996–1999.Thisnding supportsarecentstudy(Ramirez 2015 )thatfoundrainfall associations(1-monthlag)withcholeraincidence ( r2> 0.80; P value=0.000)acrossseveralcoastaldistricts inPiurain1998.Morebroadly,thendingsagreewitha waveletstudyinGhana,wherecholeraandrainfallwere synchronizedatapproximatelyalowtemporallag(ConstantindeMagnyetal. 2006 ).Theimpactsofrainfallon choleraincidencearewelldocumented(Ruiz-Morenoetal. 2007 ;Hashizumeetal. 2008 ;Rinaldoetal. 2012 ).Inlowlyingareas,suchasPiura,heavyrainscanleadtoooding incitystreets,whichmayoverwhelmseweranddrainage systems.Consequently,sewagewastecansurfaceandbe transportedviaoodwaters,whichmaycontaminatewater supplies(Currieroetal. 2001 ;Saskaietal. 2009 ).Inthecity ofPiura,seweroverowsincitystreetsarechronicenvironmentalhazards(ElTiempo 1992 ).Thus,duringthe 1997–1998ElNin ˜ o,itisplausiblethattorrentialrains exacerbatedexistinginfrastructuralproblems.Furthermore,heavyrainsmayhaveincreasedtheowandlevelof thePiuraRiver,whichinturncontributedtooodsand increasedhumanexposuretocholera. Whilerainfallassociationssuggestoodingasa mechanismforexposure,theimpactofrainfallextremeson emergingcholeraisamuchmorecomplexpathway. Table3. Pathsbandc0:MediatingEffectofLocalClimate Variables(Lag-Months)ontheNin ˜ o3.4SST(Lag-Months)and MonthlyCholeraRelationship,Piura,1996–1999. b SE t -ratio P value Outcome Cholera Pathc0:Predictor(lag) Nin ˜ o3.4SST(4)0.1700.4352.7810.008 Pathb:Mediator(lag) Rainfall(1)0.8330.08313.6690.000 Pathc0:Predictor(lag) Nin ˜ o3.4SST(4)0.2440.5193.3560.002 Pathb:Mediator(lag) RiverDischarge(0)0.7550.06110.3840.000 Pathc0:Predictor(lag) Nin ˜ o3.4SST(4)0.3071.2501.7540.086 Pathb:Mediator(lag) Tmean(6)0.4531.0722.5880.013 Table4. Pathsbandc0:MediatingEffectofLocalClimate Variables(Lag-Months)ontheNin ˜ o1+2SST(Lag-Months)and MonthlyCholeraRelationship,Piura,1996–1999. b SE. t -ratio P value Outcome Cholera Pathc’:Predictor(lag) Nin ˜ o1+2SST(6)0.1850.2373.3840.001 Pathb:Mediator(lag) Rainfall(1)0.8400.07515.3220.000 Pathc’:Predictor(lag) Nin ˜ o1+2SST(6)0.1250.3531.5280.133 Pathb:Mediator(lag) Riverdischarge(0)0.8180.0699.9970.000 Pathc’:Predictor(lag) Nin ˜ o1+2SST(6)0.7650.6115.4010.000 Pathb:Mediator(lag) Tmean(6) 0.2410.461 1.7010.096 Table5. Pathsbandc0:MediatingEffectofLocalClimate Variables(Lag-Months)onthePaitaSST(Lag-Months)and MonthlyCholeraRelationship,Piura,1996–1999. b SE t -ratio P value Outcome Cholera Pathc0:Predictor(lag) PaitaSST(6)0.2340.2934.5250.000 Pathb:Mediator(lag) Rainfall(1)0.8090.07115.6160.000 Pathc’:Predictor(lag) PaitaSST(6)0.2120.4372.7410.009 Pathb:Mediator(lag) Riverdischarge(0)0.7660.0659.9050.000 Pathc’:Predictor(lag) PaitaSST(6)0.1231.7140.4060.686 Pathb:Mediator(lag) Tmean(6)0.5861.8571.9320.06092I.J.Ram ´ rez,S.C.Grady

PAGE 11

AccordingtoRuiz-Morenoetal.( 2007 ),rainfallcanplaya dualroleincholeratransmission,wherebyexposurecan decreaseduetoadilutioneffectonbacterialconcentrations inwaterbodiesorexposurecanincreaseasoodingcontaminatewatersupplies.Ithasalsobeensuggestedthat heavyrainfallcanwashawaypredatorsof V.cholerae , whichenablesthebacteria’ssurvival,andenhancesthe potentialforexposure(Hashizumeetal. 2008 ).Morerecently,ithasbeenproposedthatincreasedriverdischarge maynotonlycontributetooodingexposurebutalso transportnutrientsintocoastalbaysthatimpactthe reproductionofvibriosandaquaticreservoirs(Jutlaetal. 2011 ).ThisisimportantbecausetheseauthorsfoundevidencethatdisputestheroleofSSTincholeraepidemics, specically,intheBayofBengal,Bangladesh.InPiura,river dischargeassociatedwiththePiuraRiverwasasignicant mediatorintheSST–cholerarelationship( b > 0.750; P value=0.00),andthus,nutrientrun-offintocoastalareas mayalsohavebeenanimportantpathwayfortransmission in1998. ThisanalysisalsofoundassociationsbetweenSSTand airtemperatureandcholerainPiura,whichsupports previousworkinLima,Peru(Francoetal. 1997 ;Speelmon etal. 2000 ;Giletal. 2004 ).AmongSSTparameters,Nin ˜ o 3.4andPaitahadthestrongesttemporallinksfrom1996to 1999,whichagreeswithRamirez( 2015 ),exceptforlag associationdifferences(0-to1-monthlag,comparedtoan estimated4-to6-monthlaginthisstudy),whichmaybe explainedbythescaleofanalysis(district)and/orperiodof study(1998).Airtemperaturewasalsosignicantly coherentwithcholerabutweakercomparedtorelationshipswithSST.Nevertheless,concurrentimpactsof anomaloustemperaturesincoastalandterrestrialenvironmentsareimportantforcholeraecology.Elevatedsea Figure6. Mediatingeffects(standardizedcoefcients)ofrainfall, riverdischarge,and TmeanintheNin ˜ o3.4SSTandcholera relationship,Piura,Peru,1996–1999. Figure7. Mediatingeffects(standardizedcoefcients)ofrainfall, riverdischarge,and TmeanintheNin ˜ o1+2SSTandcholera relationship,Piura,Peru,1996–1999. ElNin ˜ oandCholeraAssociationsinPiura,Peru93

PAGE 12

andairtemperaturescandirectlyimpactvibriosand aquaticreservoirsincoastalandinlandwaterbodies (Francoetal. 1997 ;Lippetal. 2003 ).Forexample,inLima, cholerariskwasassociatedwithairtemperaturesabove19– 20 C(Madicoetal. 1996 ;Speelmonetal. 2000 ),andfor every1 Cincrease,diarrhealdiseaseriskincreasedby * 8.0%(Checkleyetal. 2000 ;Lamaetal. 2004 ).More recently,Reyburnetal.( 2011 )showedthatcholeraincidencedoubledfollowingelevatedtemperaturesbyfour monthsinEastZanzibar.Inaddition,temperaturechanges havebeenshowntoenhancemicrobialreproductionin drinkingwaterandfoodstuffs(Tauxeetal. 1995 ).Therefore,inPiura,elevatedtemperaturesmayhaveenhanced theincubationof V.cholerae inwatersourcesinthenear shore,aswellasinmunicipalsupplies,anddrinksandfood productsfromstreetvendors,whichwereidentiedas vehiclesfortransmissionintheearly1990s(Riesetal. 1992 ). AlthoughSSTwasshowntohaveasignicantassociationwithcholera,arecentstudy(Jutlaetal. 2011 ), mentionedearlier,suggeststhattheSSTlinkmaybe coincidental.Accordingtotheauthors,theinuxof nutrientsdrivenbyriverowmayinuencetheabundance ofphytoplankton(reservoirof V . cholerae ),ratherthan SST,intheBayofBengal.Thismayexplainwhyprevious research(e.g.,Lobitzetal. 2000 )foundthatSST,whichhas beenshowntohaveaninverserelationshipwithchlorophyll(i.e.,proxyforphytoplankton),hadpositiverelationshipswithplanktonblooms(Jutlaetal. 2011 ).InPiura, thisexplanationcouldbeplausible,inpart,becauseEl Nin ˜ osnegativelyaffectbiologicalproductivityoffthecoast ofPeru.However,whileSSTmaynotaffectcholeradirectly inthiscoastalcontext,itmaystillbesignicantindirectlyif oneconsidersteleconnections.InPiura,localrainfallis stronglylinkedtoconditionsintheequatorialPacic Ocean(Lagosetal. 2008 ;Takahashi 2004 ).Rainfall,inturn, hasastrongrelationshipwithlocalriverdischarge ( r =0.78; P value=0.000).TheassociationsareparticularlyevidentduringextremeElNin ˜ os(LavadoCasimiro etal. 2012 ).Basedonthisknowledge,wehypothesizedthat ElNin ˜ o-relatedconnectionswithcholeraweremediatedby localclimate,particularlyrainfall,whichisscarceinthe region,exceptduringwarmevents.Indeed,usingmediationanalysis,wefoundthatrainfallandriverdischarge weresignicantmediatorsin1996–1999.Rainfallandriver dischargeatlagsof1-andzero-monthhadstrongmediatingeffectsacrossallSST–choleraassociations ( b > 0.750, P value < 0.05).Inaddition,itwasshownthat Tmean(lagof6months)wasamediatortoo,butonlyinthe Nin ˜ o3.4SST-cholerarelationship( b =0.453, P value=0.013). Figure8. Mediatingeffects(standardizedcoefcients)ofrainfall, riverdischarge,and TmeaninthePaitaSSTandcholerarelationship, Piura,Peru,1996–1999. Table6. Sobel’sSignicanceTestofMediation. SSTMediatorSE t -ratio P value Nin ˜ o3.4Rainfall0.7434.8600.000 Nin ˜ o3.4Riverdischarge0.7204.2670.000 Nin ˜ o3.4 Tmean1.0682.5100.012 Nin ˜ o1+2Rainfall0.4694.0340.000 Nin ˜ o1+2Riverdischarge0.4684.5640.000 Nin ˜ o1+2 Tmean0.305 1.6930.090 PaitaRainfall0.5834.4910.000 PaitaRiverdischarge0.5734.6790.000 Paita Tmean1.6191.9170.05594I.J.Ram ´ rez,S.C.Grady

PAGE 13

Takentogether,ouranalysessuggestthereisstrong evidencethatElNin ˜ oinuencedtheresurgenceofcholera inPiurain1998yetdidnotimpacttheemergencein1991. Ourinterpretationrestsonthefactthatwefoundsignicantclimate–choleralinksduringonetimeintervalassociatedwithElNin ˜ o(1997–1998).Whilecommonareasof highvariabilitywereobservedin1992–1993(inthecrosswavelet);therewasnocoherencebetweenclimatevariables andcholera.Whilethismayseemcounterintuitive,thelack ofcoherenceintheearly1990smaybeexplainedbythe orderandmagnitudeofcholeraandElNin ˜ oevents. AccordingtoRamirezetal.( 2013 ),theonsetofthe1991– 1992ElNin ˜ obeganineitherMayorNovemberof1991, dependingonthedenitionandregionchosentorepresent anevent(Nin ˜ o3.4versusNin ˜ o1+2).Both(regions)suggestthatElNin ˜ ofollowed,ratherthanprecededepidemic cholera,whichbeganinJanuary1991.Furthermore,rainfall teleconnectionswerenotobserveduntiltheaustralsummer of1992(Fig. 9 ),whichsuggeststhatElNin ˜ o-related ecosystemandweathereffectswerelikelyabsentduringthe initialepidemic.Thus,ElNin ˜ omaynothavebeenpresent toimpact V.cholerae orwaterinfrastructureintheregion. Moreover,itcouldbearguedthatthemagnitudeofElNin ˜ o (basedonSSTanomalies)wasweakerin1991–1992, comparedto1997–1998(Table 7 ),andtherefore,its capacityforimpactsonecosystemsandsocietywasless likely.LIMITATIONSThisstudyhasseveralcaveats.Onelimitationisthatour localclimatedatawerenotrepresentativeoftheentire subregion.Weusedmeteorologicalstationsinthelow-lyingcoastofPiura,whichmaynothavebeenrepresentative ofthehighlandregionsinthestudyarea.However,based oninformationabouttheepidemicsin1991and1998(Ries etal. 1992 ;Ramirez 2015 ),mostcaseswerereportedinthe coastalareasupportingthendingsofthisstudy.Specically,in1991,approximately44.0%ofallcholeracasesin theDepartmentofPiurawerereportedinthecapitalcity. Furthermore,in1998,96.2%ofcaseswerereportedinthe coastalsegmentofthesubregion.Anotherlimitationisthat wewereunabletofullyexaminetheinitialcholeraoutbreak inthewaveletanalysisbecausemuchofourdatain1991– 1992fellinsidetheconeofinuence,whichissubjectto edgeeffects—agenerallimitationofthewaveletapproach (TorrenceandCompo 1998 ).Athirdlimitationisthatwe wereunabletoincorporatesocioeconomicdataintothis analysistoassessthepotentialconfoundingeffectsofincomeandinfrastructurepovertyonclimate–cholerarelationships(Emchetal. 2010 ).Weobtained1993censusdata forPeruatthedistrictlevel,butthenexttimeperiod availablewas2007;thus,aninterpolationoftheseannual datasetswouldnotsufceforthisanalysis.Lastly,welacked datatoconstructapopulationsusceptibilityvariable(e.g., immunity)inPiuratomeasurepreviousdiseaselevels. AccordingtostudiesinBangladesh(Koelleetal. 2005 ), immunitycanimpacttheeffectsofclimateoncholera Figure9. ElNin ˜ oteleconnections(rainfallandairtemperature)in Piura,Perufrom1990–1992.SSTintheNin ˜ o3.4,1+2,andPaita coastalbayshowntoillustratethepeakofElNin ˜ o. Table7. ComparisonofSSTAnomalies(3-MonthRunning Means)Duringthe1991–1992and1997–1998ElNin ˜ os. Season1991199219971998 DJF0.4 1.8 0.4 2.3 JFM0.3 1.6 0.3 1.9 FMA0.3 1.5 0 1.5 MAM0.4 1.4 0.4 1 AMJ 0.61.20.80.5 MJJ 0.80.81.3 0 JJA 10.51.7 0.5 JAS 0.9 0.2 2 0.8 ASO 0.9 0 2.2 1 SON 1 0.1 2.4 1.1 OND 1.4 0 2.5 1.3 NDJ 1.6 0.2 2.5 1.4ElNin ˜ o(bold)andLaNin ˜ a(italics)monthsareidentiedbasedonthe ONI-index,using1971–2000baseperiodElNin ˜ oandCholeraAssociationsinPiura,Peru95

PAGE 14

incidence.Forexample,eveniffavorableclimateconditionsemerged,thelikelihoodoftransmissioncouldbelow intimesofhighimmunitybecauseoflowpopulation susceptibility.Itmay,inpart,explainwhyweonlyfounda stronglinkin1997–1998,coincidently,followinglarge outbreaksin1991–1992inPiura.CONCLUSIONInsummary,ourstudyprovidesevidencethatastrongbut transientElNin ˜ o–choleralinkinPiura,Peru,mediatedby localhydrology,existedinthelatterpartofthe1990s.To ourknowledge,thisistherststudytoexaminetheentire periodduringwhichcholerawaspresentinPiura,Peru. Furthermore,thisstudyprovidesanapproachtoestimate themediatingeffectsoflocalclimateonapotentialEl Nin ˜ o–cholerarelationship.Futureresearchshouldexamine furtherthemediatingeffectsofhydrologyontheElNin ˜ o– choleralink.Forexample,investigatingriverdischargein relationshiptoSSTmayprovidesupportforadual mechanismforSST(i.e.,impactsonvibrioecologyand inuenceonlocalweather).Inaddition,wefoundno evidenceofanElNin ˜ olinkintheearlierpartofthedecade. ThisisimportantbecauseitprovidessupportthatElNin ˜ o maynothaveprecipitatedcholeraemergenceinPeru (Ramirezetal. 2013 ).Still,beforetheseresultscanbe conclusive,temporalexaminationsofElNin ˜ olinksinother choleraepicenters(e.g.,ChancayandLima)shouldbe undertaken.Also,itwillbeimportantforthesestudiesto alsoexplorewhyElNin ˜ owasinuentialin1997–1998and notin1991–1992.AccordingtoCapotondietal.( 2015 ), impactsare‘‘highlysensitive’’tothevariabilityofENSO characteristicsfromoneeventtoanother.Thus,bycomparingthetwoevents,wemaylearnhowthediversityofEl Nin ˜ osimpactcholeratransmission.Moreover,future studiesshouldincorporateotherexplanatoryfactors,e.g., humanimportation,herdimmunity,socioeconomicvulnerability,publichealtheducation,oraconvergenceof variables,includingclimate. Forpublichealthprogramming,thisstudyhighlights thepotentialutilityofglobaltolocalhydro-meteorological informationfordiseaseprevention.Inparticular,itmay informexistingeffortsthatutilizeElNin ˜ omonitoringto mobilizehealthpersonnelandresourcesinanticipationof extremeweather(Sandoval 1999 ).Atthesametime,it suggestscautionandcarefulattentiontoElNin ˜ o-related characteristicsindecision-making.WhileElNin ˜ omay provideanopportunityforearlywarning,itsdevelopment mayvaryinintensityandimpacts(Glantz 1991 ),asmentionedearlier;andthussotoomayitsinuenceonweather anddiseaseecology.Nevertheless,concernsabout reemergingcholeraintheregion,aswellasthepotential impactsofachangingclimate,warrantabettercomprehensionofclimatedynamicstoimprovecholerapreparednessduringfutureclimate-relatedextremes.ACKNOWLEDGMENTSTheauthorswouldliketothanktheDepartmentof Geography,MichiganStateUniversityforthenancial supportfordatacollectioninPeru.WealsothankTheNew Schoolforprovidingthespaceandfundingtocompletethe manuscript(ResearchFacultyFundandReNewSchool Project14kGrantAward).Wearealsogratefultoour Peruviancollaborators,includingIng.NormaOrdinolaand Ing.RodolfoRodriguez,UniversityofPiura,Ing.Grover Otero,ProyectoChira-PiuraandDr.ElsaGalarza,UniversityofPacic,aswellastheDepartmentsofEpidemiology attheMinistriesofHealth,andtheInstituteforStatistics andInformationinLimaandPiura,Peru.APPENDIXSeeFig. 10 . 96I.J.Ram ´ rez,S.C.Grady

PAGE 15

Figure10.a i WaveletcoherenceanalysesbetweenSST andlocalclimatevariables: a Nin ˜ o3.4SSTanomalyand rainfall(square-roottransformed); b Nin ˜ o1+2SST anomalyandrainfall (square-roottransformed); c PaitaSSTanomalyand rainfall(square-roottransformed); d Nin ˜ o3.4SST anomalyandriverdischarge (square-roottransformed); e Nin ˜ o1+2SSTanomaly andriverdischarge(squareroottransformed); f Paita SSTanomalyandriverdischarge(square-roottransformed); g Nin ˜ o3.4SST anomalyandairtemperature( Tmean)anomaly;h Nin ˜ o1+2SSTanomalyand airtemperature( Tmean) anomaly;and i PaitaSST anomalyandairtemperature( Tmean)anomaly.The waveletcoherenceanalysisis denotedbyperiod(scaleby year)andacrosstimeintervals.Thecolorcodeshows coherencevaluesthatincreasefrom darkblue (low) to darkred (high).The direction(phase)ofrelationshipsisindicatedby arrows ,assuch:up(climate lags);down(climateleads); right(climate-cholerainphase);andleft(climatecholeraoutofphase).Statisticalsignicance(95.0% condencelevel)isindicatedbyareaswithin thick blackoutlines .The black curve delimitstheconeof inuence(COI),aregion inuencedbyedgeeffects. ElNin ˜ oandCholeraAssociationsinPiura,Peru97

PAGE 16

REFERENCESBarronRM,KennyDA(1986)Themoderatormediatorvariable distinctioninsocialpsychologicalresearch:conceptual,strategic,andstatisticalconsiderations. JournalofPersonalityand SocialPsychology 51:1173–1182 CapotondiA,WittenbergAT,NewmanM,DiLorenzoE,YuJ, BraconnotP,etal.(2015)UnderstandingENSOdiversity. BulletinoftheAmericanMeteorologicalSociety 96:921–938. doi: 10.1175/BAMS-D-13-00117.1 CashBA,Rodo ´ X,KinterJL(2008)LinksbetweentropicalPacic SSTandcholeraincidenceinBangladesh:roleoftheeasternand centraltropicalPacic. JournalofClimate 21:4647–4663 CazellesB,ChavezM,ConstantindeMagnyG,GueganJ,HalesS (2007)Time-dependentspectralanalysisofepidemiological time-serieswithwavelets. TheJournaloftheRoyalSociety Interface 4:625–636 Cazelles,B,Chavez,M,McMichael,AJ,Hales,S(2005)NonstationaryinuencesofElNin ˜ oonthesynchronousdengueepidemicsinThailand. PublicLibraryofScience 2:e106.doi: 10. 1371/journal.pmed.0020106 .AccessedFebruary1,2015 Chaves,LF,Pascual,M(2006)Climatecyclesandforecastsof CutaneousLeishmaniasis,anonstationaryvector-bornedisease. PublicLibraryofScience,3:e295.doi: 10.1371/journal.pmed. 0030295 .AccessedFebruary1,2015 CheckleyW,EpsteinLD,GilmanRH,FigueroaD,CamaRI,Patz JA,etal.(2000)EffectofElNin ˜ oandambienttemperatureon hospitaladmissionsfordiarrhoealdiseasesinPeruvianchildren. TheLancet 355:442–450 ColwellRR(1996)Globalclimateandinfectiousdisease:the choleraparadigm. Science 274:2025–2031 ConstantindeMagnyG,CazellesB,GueganJF(2006)Cholera threattohumansinGhanaisinuencedbybothglobaland regionalclimatevariability. EcoHealth 3:223–231 CurrieroFC,PatzJA,RoseJB,LeleS(2001)TheAssociation BetweenextremeprecipitationandwaterbornediseaseoutbreaksintheUnitedStates,1948–1994. AmericanJournalof PublicHealth 91:1194–1199 ElTiempo(TheTimes)(1992)Criticasituacio ´ nafrontanlos distritosporlluvias(Districtsfaceacriticalsituationcausedby rains).18March,Piura(inSpanish) EmchM,YunusM,EscamillaV,FeldackerC,AliM(2010)Local populationandregionalenvironmentaldriversofcholerain Bangladesh. EnvironmentalHealth 9. http://www.ehjournal.net/ content/9/1/2 .AccessedFeb15,2015 EpsteinPR,FordTE,ColwellRR(1993)Healthandclimate change:marineecosystems. TheLancet 342:1216–1219 FrancoAA,etal.(1997)CholerainLima,Peru,correlateswith priorisolationof Vibriocholerae fromtheEnvironment. AmericanJournalofEpidemiology 146:1067–1075 FrazierPA,TixAP,BarronKE(2004)Testingmoderatorand mediatoreffectsincounselingpsychologyresearch. Journalof CounselingPsychology 51:115–134 GilAI,LouisVR,RiveraING,LippE,HuqA,LanataCF,etal. (2004)Occurrencean ddistributionof Vibriocholerae inthecoastal environmentofPeru. EnvironmentalMicrobiology 6:699–706 GlantzMH(1991)Introduction.In: TeleconnectionsLinking WorldwideClimateAnomalies ,GlantzMH,KatzRW,Nicholls N(editors),NewYork:CambridgeUniversityPress,pp2–12 GradySC,RamirezIJ(2008)Mediatingmedicalriskfactorsintheresidentialsegregationandlowbirthweightrelationshipbyrace inNewYorkCity. HealthandPlace 14:661–677 GrinstedA,MooreJC,JevrejevaS(2004)Applicationofthecross wavelettransformandwaveletcoherencetogeophysicaltime series. NonlinearProcessesinGeophysics 11:561–566 GrinstedA,MooreJC,JevrejevaS(2008)Matlabpackagefor performingcross-waveletandwaveletcoherence. http://noc.ac. uk/using-science/crosswavelet-wavelet-coherence .AccessedFeb 1,2015 HashizumeM,ArmstrongB,HajatS,WagatsumaY,FaruqueAS, HayashiT,etal.(2008)Theeffectofrainfallontheincidenceof cholerainBangladesh. Epidemiology 19:103–110 HashizumeM,ChavesLF,FarunqueASG,YunusMD,Streateld K,KazuhikoM(2013)AdifferentialeffectofIndianOcean DipoleandElNinooncholeradynamicsinBangladesh. PLoS ONE 8(3):e60001.doi: 10.1371/journal.pone.0060001 IBMSPSS(2015)Predictiveanalyticssoftwareandsolutions. http://01.ib.com/software/analytics/spss .AccessedMar22,2015 InstitutoNacionaldeEstadisticaeInformatica(InstituteofNationalStatisticsandInformation)(INEI)(2000)LasEstadisticas VitalesenLosDistritosdelPeru(VitalStatisticsattheDistrictlevelinPeru. http://www.inei.gob.pe/media/MenuRecursivo/ publicaciones_digitales/Est/Lib0397/Libro.pdf .AccessedFeb16, 2015 JutlaA,WhitcombeE,HasanN,HaleyB,AkandaA,HuqA,etal. (2013)Environmentalfactorsinuencingepidemiccholera. The AmericanJournalofTropicalMedicineandHygiene 89:597–607 JutlaAS,AkandaAS,GrifthsJK,ColwellRR,IslamS(2011) Warmingoceans,phytoplankton,andriverdischarge:implicationsforcholeraoutbreaks. AmericanJournalofTropicalMedicineandHygiene 85:303–308 KoelleK,RodoX,PascualM,YunusM,MostafaG(2005)Refractoryperiodsandclimateforcingincholeradynamics. Nature 436:696–700 LagosP,SilvaY,NicklE,MosqueraK(2008)ElNin ˜ o-related precipitationvariabilityinPeru ´ . AdvancesinGeosciences 14:231–237. http://www.adv-geosci.net/14/231/2008/ .Accessed July20,2013 LamaJR,SeasCR,Leon-BaruaR,GotuzzoE,SackRB(2004) Environmentaltemperature,cholera,andacutediarrheain adultsinLima,Peru. JournalofHealthandPopulationNutrition 22:399–403 LavadoCasimiroWS,RonchailJ,LabatD,EspinozaJC,GuyotJL (2012)Basin-scaleanalysisofrainfallandrun-offinPeru (1969–2004):Pacic,TiticacaandAmazonasdrainages. HydrologicalSciencesJournal 57:625–642. http://www.igp.gob.pe/ igp/images/documents/comunicaciones/publicaciones/2012/per utrends_iahs_2012.pdf .AccessedFeb15,2015 LippEK,etal.(2003)Directdetectionof Vibriocholerae andctxA inPeruviancoastalwaterandplanktonbyPCR. Appliedand EnvironmentalMicrobiology 69:3676–3680 LobitzB,BeckL,HuqA,WoodB,FuchsG,FaruqueASG,Colwell RR(2000)Climateandinfectiousdisease:useofremotesensing fordetectionofV.choleraebyindirectmeasurement. ProceedingsoftheNationalAcademyofSciencesoftheUnitedStatesof America 97:1438–1443 MadicoGW,CheckleyW,GilmanRH,BravoN,CabreraL, CalderonM,etal.(1996)Activesurveillancefor Vibriocholerae 01andvibriophagesinsewagewaterasapotentialtoolto98I.J.Ram ´ rez,S.C.Grady

PAGE 17

predictcholeraoutbreaks. JournalofClinicalMicrobiology 34:2968–2972 MinisteriodeSalud(MinistryofHealth)(2005) Protocolosdela vigilanciaepidemiologica (ProtocolsforEpidemiological Surveillance),Part1.Lima:DepartmentofEpidemiology, MinistryofHealth(inSpanish) MooreSM,ShannonKL,ZelayaCE,AzmanAS,LesslerJ(2014) EpidemicriskfromcholeraintroductionsintoMexico. PLOS CurrentsOutbreaks ,Edition1.DOI: 10.1371/currents.outbreaks. c04478c7fbd9854ef6ba923cc81eb799 NgEKW,KwokTW(2012)Thesoftwareforthebias-rectied waveletpowerspectrum,partialwaveletcoherenceandmultiple waveletcoherence. http://www.cityu.edu.hk/gcacic/wavelet .AccessedSep10,2014 NOAA(2015)Coldandwarmepisodesbyseason.ClimatePredictionCenter. http://www.cpc.ncep.noaa.gov/products/analy sis_monitoring/ensostuff/ensoyears.shtml .AccessedFeb16,2015 PAEN/GTZ(2003) DiagnosticozonabajaCuencaRioPiuracon enfoquedegestiondelriesgo (VulnerabilityAssessmentofthe LowerZoneoftheRiverPiuraWatershed).GermanAgencyfor TechnicalCooperation(Unpublished,inSpanish) PAHO(2008)ThenumberofcholeracasesintheAmericas,1990– 2008. http://www.paho.org/English/AD/DPC/CD/cholera-19902008.pdf .AccessedFeb10,2015 PanAmericanHealthOrganization(PAHO)(2014)Atlasof choleraoutbreakinLaHispaniola,2010–2014. http://new.paho. org/hq/images/Atlas_IHR/Ch oleraHispaniola/atlas.html .Accessed Feb16,2015 PascualM,BoumaMJ,DobsonAP(2002)Choleraandclimate: revisitingthequantitativeevidence. MicrobesandInfection 4:237–245 PascualM,RodoX,EllnerSP,ColwellRR,BoumaMJ(2000) CholeradynamicsandElNin ˜ o-SouthernOscillation. Science 289:1766–1769 PreacherKJ,HayesAF(2004)SPSSandSASproceduresfor estimatingindirecteffectsinsimplemediationmodels. Behavior ResearchMethods,Instruments,&Computers 36:717–731 RamirezIJ(2015)CholeraresurgenceinPiura,Peru:examining climateassociationsduringthe1997-98ElNin ˜ o. GeoJournal 80:129–143.doi: 10.1007/s10708-014-9541-2 RamirezIJ,GradyS,GlantzMH(2013)ReexaminingElNin ˜ oand cholerainPeru:aclimateaffairsapproach. Weather,Climate andSociety 5:148–161 ReyburnR,KimDR,EmchM,KhatibA,vonSeidleinL,AliM (2011)Climatevariabilityandtheoutbreaksofcholerain Zanzibar,EastAfrica:atimeseriesanalysis. TheAmerican JournalofTropicalMedicineandHygiene 84:862–869. doi: 10.4269/ajtmh.2011.10-0277 RiesAA,VugiaDJ,BeingoleaL,PalaciosAM,VasquezE,Wells JG,etal.(1992)CholerainPiura,Peru ´ :amodernurbanepidemic. TheJournalofInfectiousDiseases 166:1429–1433 RinaldoA,BertuzzoE,MariL,RighettoL,BlokeschM,GattoM, etal.(2012)Reassessmentofthe2010-2011Haiticholeraoutbreakandrainfall-drivenmultiseasonprojections. Proceedingsof theNationalAcademyofSciencesoftheUnitedStatesofAmerica 109:6602–6607.doi: 10.1073/pnas.1203333109 Ruiz-MorenoD,PascualM,BoumaM,DobsonA,CashB(2007) CholeraseasonalityinMadras(1901-1940):dualroleforrainfall inendemicandepidemicregions. EcoHealth 4:52–62 SandovalPS(1999) Evaluacio ´ ndeDan˜ osyAccionesdelFeno ´ meno ElNin ˜ o (EvaluationofDamagesCausedbytheElNin ˜ oPhenomenonandActionsTaken).Peru:OcinadePlanicacio ´ n, Direccio ´ ndeSaludRegional(PlanningOfce,RegionalHealth Post)(inSpanish) SaskaiS,SuzukiH,YasuyukiF,YoshinariK,MeetwellC(2009) ImpactofdrainagenetworksoncholeraoutbreaksinLusaka, Zambia. AmericanJournalofPublicHealth 99:1982–1987 SpeelmonEC,CheckleyW,GilmanRH,PatzJ,CalderonM, MangaS(2000)CholeraincidenceandElNin ˜ o-relatedhigher ambienttemperature. JournaloftheAmericanMedicalAssociation 283:3072–3074 TakahashiK(2004)TheatmosphericcirculationassociatedwithextremerainfalleventsinPiura,Peru,duringthe1997–98and2002El Nin ˜ oevents. AnnalesGeophysicae 22:3917–3926. http://www.anngeophys.net/22/3917/2004/angeo-22-3917-2004.pdf .AccessedJuly 20,2013 Tauxe,R,Mintz,ED,Quick,RE(1995)Epidemiccholerainthe NewWorld:translatingeldepidemiologyintonewstrategies. EmergingInfectiousDiseases,1.Available: http://wwwnc.cdc. gov/eid/article/1/4/95-0408 [accessedFebruary15,2015] TorrenceC,CompoGP(1998)Apracticalguidetowavelet analysis. BulletinoftheAmericanMeteorologicalSociety 79:61–78 ElNin ˜ oandCholeraAssociationsinPiura,Peru99