Extreme weather and economic well-being in rural Mozambique

MISSING IMAGE

Material Information

Title:
Extreme weather and economic well-being in rural Mozambique
Series Title:
Natural Hazards
Physical Description:
Journal Article
Creator:
Matyas, Corene
Publication Date:

Notes

Abstract:
Societies dependent on rain-fed agriculture are highly vulnerable to weather extremes; thus, linkages between rainfall variability and economic well-being merit close attention. The hypothesis of this paper is that rainfall patterns impact changes in income within our study region of central and northern Mozambique. Utilizing satellite-based estimates of rainfall analyzed within a GIS, we establish a 12-year rainfall climatology and calculate monthly rainfall anomalies for 419 villages during three growing seasons. We also approximate storm-total rainfall from tropical cyclones entering the Mozambique Channel. Hierarchical cluster analysis groups the villages according to the monthly rainfall anomalies and rainfall received from Cyclones Delfina and Japhet. Then, using data from the National Agricultural Survey of Mozambique conducted in 2002 and 2005, we relate rainfall and change in income through the calculation of Pearson’s correlation coefficients and independent-samples t tests using village-groups produced by the cluster analysis. We find that no season closely approximates the 12-year climatology and that rainfall varied among the three seasons. Although most villages experience income declines, those affected by Delfina exhibit the worst economic performance, indicating that heavy rainfall from some tropical cyclones can have long-lasting negative effects on income. Additionally, receiving above-normal rainfall may hinder economic well-being more than below-normal rainfall. Our study identifies patterns in sub-national rainfall variability and economic well-being that enable a more detailed understanding of weather-related effects on socio-economic outcomes.
Acquisition:
Collected for University of Florida's Institutional Repository by the UFIR Self-Submittal tool. Submitted by Corene Matyas.
Publication Status:
Published

Record Information

Source Institution:
University of Florida Institutional Repository
Holding Location:
University of Florida
Rights Management:
All rights reserved by the submitter.
System ID:
IR00003676:00001


This item is only available as the following downloads:


Full Text

PAGE 1

ORIGINALPAPERExtremeweatherandeconomicwell-beinginrural MozambiqueCoreneJ.Matyas€JulieA.SilvaReceived:15July2011/Accepted:8December2011/Publishedonline:20December2011 SpringerScience+BusinessMediaB.V.2011Abstract Societiesdependentonrain-fedagriculturearehighlyvulnerabletoweather extremes;thus,linkagesbetweenrainfallvariabilityandeconomicwell-beingmeritclose attention.Thehypothesisofthispaperisthatrainfallpatternsimpactchangesinincome withinourstudyregionofcentralandnorthernMozambique.Utilizingsatellite-based estimatesofrainfallanalyzedwithinaGIS,weestablisha12-yearrainfallclimatologyand calculatemonthlyrainfallanomaliesfor419villagesduringthreegrowingseasons.We alsoapproximatestorm-totalrainfallfromtropicalcyclonesenteringtheMozambique Channel.Hierarchicalclusteranalysisgroupsthevillagesaccordingtothemonthlyrainfall anomaliesandrainfallreceivedfromCyclonesDelnaandJaphet.Then,usingdatafrom theNationalAgriculturalSurveyofMozambiqueconductedin2002and2005,werelate rainfallandchangeinincomethroughthecalculationofPearson'scorrelationcoefcients andindependent-samples t testsusingvillage-groupsproducedbytheclusteranalysis.We ndthatnoseasoncloselyapproximatesthe12-yearclimatologyandthatrainfallvaried amongthethreeseasons.Althoughmostvillagesexperienceincomedeclines,those affectedbyDelnaexhibittheworsteconomicperformance,indicatingthatheavyrainfall fromsometropicalcyclonescanhavelong-lastingnegativeeffectsonincome.Additionally,receivingabove-normalrainfallmayhindereconomicwell-beingmorethan below-normalrainfall.Ourstudyidentiespatternsinsub-nationalrainfallvariabilityand economicwell-beingthatenableamoredetailedunderstandingofweather-relatedeffects onsocio-economicoutcomes. Keywords Climatevariability Rainfall Ruraldevelopment Africa Mozambique Tropicalcyclones C.J.Matyas( & ) DepartmentofGeography,UniversityofFlorida,3141TurlingtonHall,Box117315, Gainesville,FL32611,USA e-mail:matyas@u.edu J.A.Silva DepartmentofGeography,UniversityofMaryland,CollegePark,MD,USA123NatHazards(2013)66:3149 DOI10.1007/s11069-011-0064-6

PAGE 2

1Introduction Thedevelopmentliteraturehasincreasinglyemphasizedtheroleofclimatevariabilityin shapingvulnerabilityinruralcommunities,particularlyinsouthernAfrica(O'Brienand Vogel 2003 ;LeichenckoandO'Brien 2002 ).Growinginterestintheinteractionsofeconomicandclimateriskillustratestheneedforbetterunderstandingsofhowextreme weatherimpactspeopleinthedevelopingworld(Mirza 2003 ;SchipperandPelling 2006 ). PreviousresearchershaveestablishedthatsouthernAfricaexperienceshighvariabilityin rainfallonaninterannualbasis(Cooketal. 2004 ;Reasonetal. 2005 ;RouaultandRichard 2005 ;UsmanandReason 2004 ).Climate-changescenariospredicta1020%decreasein rainfalloversouthernAfricaby2030(Lobelletal. 2008 ;SitholeandMurewi 2009 ),thus increasingthevulnerabilityoffarmersrelyingonrain-fedagriculture(BrownandFunk 2008 ).CountriesinsouthernAfricaalsoconfrontmanycriticalobstaclestoeconomic development,includinghighpovertyrates,corruption,weakinfrastructure,andhighrates ofdisease(WorldBank 2010 ). SeveralelementsmakeMozambiqueanappropriatecasestudyforinvestigatinglinkages betweenweatherandeconomicwell-beinginruralareas.ThemajorityofruralMozambicansarepoor,with54%livingbelowthepovertylinein2005and80%ofthepeoplerelying onrain-fedagricultureforsubsistence.Evidencealsosuggeststhatpovertyhasincreasedin therecentdecadesduetofrequentnaturalhazards,lowagriculturalproductivity,lackof basicinfrastructure,poorhealth,andmarketfailures(HallandYoung 1997 ;Mittleman 2000 ;Pitcher 2002 ;Sheldon 2002 ;HanlonandSmart 2008 ;CunguaraandHanlon 2010 ). Povertylimitstheabilityofsmallholderfarmerstomitigatethenegativeeffectsofextreme weatherevents(Mirza 2003 ;Hahnetal. 2009 ;EriksenandSilva 2009 ).Eightmajoroods andsevenmajordroughtshaveaffectedthecountryduringthepast30years(Klinmanand Reason 2008 ),andsucheventshavebeenlinkedtosignicantnegativeimpactsoncrop production(UsmanandReason 2004 ).Inaddition,Mozambiquehasexperienceddamaging tropicalcyclonesthathavehadlong-lastingimpactsonsociety,includingthebreakdownof informalsocialsupportnetworks(BrouwerandNhassengo 2006 ;ChristieandHanlon 2001 ). AsresearcherspredictthatsouthernAfricawillexperienceanincreaseinclimatevariability, extremeeventsintheregionarelikelytocontinue(IPCC 2007 ). Ourhypothesisisthatrainfallpatternsimpactchangesinhouseholdincomeinourstudy region.Asarststeptowardlinkingrainfallvariabilitytoeconomicwell-beingin Mozambique,thisstudyexaminesrainfallpatternsduringthethreegrowingseasonsthat occurredbetweentwonationalsocio-economicsurveysofruralhouseholdsconductedin 2002and2005.WefocusonvillagesinthecentralandnorthernregionsofMozambiqueas theyshareacommongrowingseason.Asmostclimate-scalestudiesofprecipitationhave beenforthelargerregionofsouthernAfrica(e.g.,Juryetal. 2004 ;Shongweetal. 2009 ; UsmanandReason 2004 ),wedevelopasatellite-basedrainfallclimatologyandexplore monthlyrainfallpatternsatthesub-nationallevel.Extremedroughtoccurredinsouthern Africafrom2002to2005(Matheretal. 2008 ;RouaultandRichard 2005 ),havingmajor impactsonagricultureinMozambiqueasreportedbytheFoodandAgricultureOrganizationoftheUnitedNations(FAO 2003 2004 2005 ).Additionally,twotropicalcyclones causeddifferentrainfallpatternsinourstudyregionduringthisperiod(Kadomura 2005 ). Assuch,ourstudyseekstodeterminewhetherreceivingbelow-normalorabove-normal rainfallexhibitsahighercorrelationwiththechangesinruralhouseholdincomeoverthe studyperiod. Ourstudyregioncontained419villagesforwhichthesocio-economicdataareavailable.Satellite-basedrainfallestimatesareanalyzedwithinaGIStoestablisharainfall 32NatHazards(2013)66:3149123

PAGE 3

climatology,determinethepercentofnormalrainfalloccurringwithineachofthestudy months,andcalculaterainfalltotalsthatresultedfromthepassageoftropicalcyclonesfor eachvillage.Correlationcoefcientsarecalculatedtodemonstratetheextenttowhich rainfallpatternsdeviatedfromnormal,andtorelatemonthlyrainfallandtropicalcyclone rainfalltotalstochangesinincome.Tofurtherexplorelinkagesbetweenincomeand rainfall,weperformahierarchicalclusteranalysistogroupvillagesaccordingtothe rainfalltheyreceivedacrossthestudyperiod.Themeanchangesinincomewithinthese groupsarecomparedusingindependentsamples t teststodeterminewhetheranyofthese groupsexperiencedchangesinincomethataresignicantlydifferentfromtheother groups.Ourndingsidentifypatternsinsub-nationalrainfallvariabilityandeconomic well-beingthatenableamoredetailedanalysisofweather-relatedeffectsonsocio-economicoutcomes. 2RainfallclimatologyofcentralandnorthernMozambique InordertounderstandrainfallpatternsatthesubnationallevelinMozambique,weplace theminthewidercontextofthephysicalmechanismsdrivingprecipitationintheregion.In northernandcentralMozambique(Fig. 1 ),thegrowingseasonspansNovemberApril Fig.1 MapofMozambiqueshowingthelocationsofthestudyvillages,elevation,andtracksofalltropical cyclonesduringthestudyperiod NatHazards(2013)66:314933123

PAGE 4

(Tadrossetal. 2009 ),andinterannualvariationsinrainfalltotalsarehigh(Cooketal. 2004 ;Fauchereauetal. 2003 ).Rainfalloccursinassociationwiththemigrationofthe intertropicalconvergencezone(ITCZ)toalatitudeofapproximately20 S(Washington andPreston 2006 ),withmoistureadvectionprimarilyfromthesouthwesternIndianOcean (Reason 2007 ;Thomasetal. 2007 ).Tropicaltemperaturetroughs(TTTs)andtheirassociatednorthwest-southeastorientedcloudbandslinktheAngolalowwithwesterlydisturbancesinthemiddlelatitudesofthesouthernhemisphere(Reason 2007 ;Shongweetal. 2009 ;UsmanandReason 2004 ).AlthoughTTTsoccurinfrequentlyandcandevelopin differentlocationsoverthecontinentonbothintra-andinter-annualtimescales,theyare responsibleforthemajorityoftherainfallduringtheaustralsummer(Cooketal. 2004 ; UsmanandReason 2004 ;WashingtonandTodd 1999 ).Topographicascentcanalso enhancerainfalloverthenorthwesternportionofMozambique(Thomasetal. 2007 ; Wisner 1979 )(Fig. 1 ).Duringdrierseasons,anticyclonicconditionsprevailastheITCZ remainsnorthof20 S(Cooketal. 2004 ;JuryandPathack 1993 ). Multipleatmosphericteleconnectionscontributetotheinterannualvariabilityofrainfall overMozambique,includingtheElNino-SouthernOscillation(ENSO),theIndianOcean dipole(IOD),andthesubtropicalIndianOceandipole(SIOD).Researchersagreethat ENSOeventsstronglyinuencerainfalloversouthernAfrica(Juryetal. 2004 ;Reasonand Jagadheesha 2005 ;RouaultandRichard 2005 ;UsmanandReason 2004 ;Watterson 2009 ). WhenLaNinaconditionsprevail,TTTsanchoroverthecontinentnear25 Etoproduce rainfalloverMozambique(Juryetal. 2004 ).DuringElNinoevents,TTTsshiftto approximately55 Esothatsubsidenceassociatedwithanomaloushighpressureandalack ofrainfalldominatessoutheasternAfrica(Juryetal. 2004 ;Reason 2007 ).Anomaloussea surfacetemperatures(SSTs)inthetropicalIndianOceanhavealsobeenlinkedtorainfall variationsinsouthernAfrica.WhentheIODispositive,SSTsareanomalouslyhighinthe equatorialwesternIndianOcean,leadingtoenhancedconvectioninthisregion,which includescoastalnorthernMozambique.DuringpositiveIODmonths,subsidenceis enhancedelsewhereoversouthernAfrica(Manatsaetal. 2008 ;Richardetal. 2001 ).The positivephaseoftheSIODindicatesthatanomalouslywarmSSTsarepresentsouthof Madagascar.EvaporationfromtheoceanandmoistureadvectionintoMozambiquefrom thesoutheastareenhanced(BeheraandYamagata 2001 ;HansingoandReason 2009 )so thatpositivevaluesoftheSIODareassociatedwithincreasedrainfallovermostof Mozambique,whilenegativevaluesofthisindexareassociatedwithdryness.Boththe IODandtheSIODhavebeenshowntooperateindependentlyofENSO. TropicalcyclonesprovideanothersourceofrainfallforMozambique.Rainfallfrom thesesystemscanalleviatedroughtconditions(Angel 2006 ;Sugg 1968 )orcanproduce severeooding(Reason 2007 ;ReasonandKeibel 2004 ;Vitartetal. 2003 ).Atropicalcyclonecancauseextensiveoodinginaregioneventhoughitscirculationcenterdoesnot passdirectlyoverhead,asitsrainshieldscanextendhundredsofkilometersoutwardfrom thecenterofcirculation(Lonfatetal. 2007 ;Matyas 2010 ).Accordingtocyclonetrackdata obtainedfromtheJointTyphoonWarningCenter(JTWC)( http://www.usno.navy.mil/ NOOC/nmfc-ph/RSS/jtwc/best_tracks/shindex.html )andplottedwithinaGIS,87cyclones havetrackedwithin100kmofMozambiqueduringthe60-yearperiod19492008.Of these,45havemadelandfall.Thus,onaveragethecountrycouldexperiencerainfallfroma cyclone12timesperyear.LikethepositioningofTTTs,thetracksoftropicalcyclones areassociatedwithatmosphericteleconnections.DuringLaNinaevents,and/orwhenthe SIODisinapositivephase,cyclonestendtomoveinzonal(east-to-west)patternsacross theSWIO,increasingtheprobabilityofaffectingMozambique.DuringElNinoyearsand/ orwhentheSIODisnegative,cyclonesrecurvetowardthesoutheastbeforeenteringthe 34NatHazards(2013)66:3149123

PAGE 5

MozambiqueChannel(AshandMatyas 2012 ;Vitartetal. 2003 ),solandfallsin Mozambiquearelessfrequentwheneitheroftheseconditionsprevail. 3Rainfallanalysesandresults 3.1Methodofanalysis Weconductourrainfallanalysesatmonthlyandseasonalscalestoenhancecomparability withtheFAO'sreportedrainfallimpactsonagricultureinMozambique.Asraingaugedata withhighspatialandtemporalresolutionaresparseovermostofAfrica(Adeyewaand Nakamura 2003 )andparticularlywithinMozambique,weestimaterainfalltotalsthrough ananalysisoftheTropicalRainfallMeasuringMission(TRMM)'s3B43product.Ouruse ofthesedatatoestimaterainfallisjustiedasAdeyewaandNakamura( 2003 )andDinku etal.( 2007 )foundthattheTRMM3B43productexhibitedahighcorrelationwithrain gaugemeasurementsoverseveralregionswithinAfrica.Thealgorithmthatcreatesthis productcombinespassivemicrowaveandinfraredsatellite-baseddatawithmeasurements fromraingaugestoproducerainfalltotals.Dataareavailableataspatialresolutionof 0.25 9 0.25 forlatitudes50 N50 S,andareavailablemonthlybeginninginJanuary, 1998(Huffmanetal. 2007 ).The3B43dataspanning19982009areacquiredusingthe GES-DISCInteractiveOnlineVisualizationANdaNalysisInfrastructure(Giovanni)as partoftheNASA'sGoddardEarthSciences(GES)DataandInformationServicesCenter (DISC).MonthlyrainfalltotalsforNovemberAprilareobtainedforaregionspanning 9 30 Sand26 46 E.However,weeliminateAprilfromtheanalysisaswecannotmatch thesatellite-derivedrainfallduringApril2004toraingaugedatapublishedbytheFAO ( 2004 )andGodfreyetal.( 2005 ). First,weconstructa12-yearrainfallclimatologybyimportingthemonthlyTRMMestimatedrainfalldataintoaGISandinterpolatingthedatausingordinaryspherical kriging.Weassignvaluestoeachofthe419villages,thenaddthevaluesforeachyearand dividebytwelvetoobtaintheclimatologicalaverage.Next,weexaminetheactualrainfall occurringduringthestudyperiod.Wedividetherainfalloccurringeachmonthbyits climatologicalvaluetoobtainthepercentageofnormalrainfallateachvillagetofacilitate comparisonsacrossareasthatnormallyreceivedifferentrainfallamounts.Wealsosum thesepercentagesanddividebyvetoobtaintheaveragepercentageofnormalrainfall receivedNovembertoMarchforeachseason.Season1spans20022003,whileSeasons2 and3cover20032004and20042005,respectively. Toassesstherainfallcontributedbytropicalcyclonesduringourstudyperiod,we importthecyclonetrackdataobtainedfromtheJTWCintotheGIS.Allcycloneswithin theMozambiqueChannel(Fig. 1 )areextractedforfurtheranalysis.Rainfalldataare obtainedfromtheTRMM3B42V6deriveddataset( http://disc2.nascom.nasa.gov/ Giovanni/tovas/TRMM_V6.3B42_daily.shtml ),whichissimilartotheaforementioned 3B43productbutcontainsdailyrainfallvalues.WeimporttheTRMMdataintotheGIS andperformordinarysphericalkriging,thenaddthedailyrainfallvaluestodeterminethe storm-totalrainfallproducedbyeachtropicalcyclone.Ouranalysisrevealsthatonlytwo cyclonesproducedmorethan150mmofrainfallaffectingatleast30villageswithinour studyregionandstudyperiod.CycloneDelnaanditsremnantcirculationproduced rainfallovertheregionfrom31December20026January2003,whilerainfallfrom CycloneJaphetproducedrainfalloverthestudyregion15March2003.Asweselect 150mmasathresholdforvillagesreceivingveryhighrainfalltotalsfromcyclones,wedo NatHazards(2013)66:314935123

PAGE 6

notdiscussCycloneAtang,whichmadelandfallinTanzaniabutproduced50140mmof rainfallinsevenMozambicanvillagesneartheTanzanianborderinNovember2002. Weutilizetwostatisticaltechniquestoexplorethespatialandtemporalpatternsof rainfall.Afterconrmingthatallvariablesaredistributednormally,Pearson'scorrelation coefcientsarecalculatedbetweentheclimatologicalandactualrainfallvaluesforeach monthaswellaslatitudeandlongitudetocharacterizetheextenttowhichrainfalldiffered fromnormal.Toexplorethespatialpatternsofthenormalandactualrainfall,weperform twoseparatehierarchicalclusteranalyses.Therstemploysclimatologicalrainfallvalues ineachstudymonth,whilethesecondutilizesthepercentageofnormalrainfallforall15 studymonths.Bothanalysesareperformedemphasizingbetween-groupslinkagewith intervalsdeterminedbysquaredEuclideandistance.Optimalclusteringsolutionsare selectedbaseduponthespatialpatternstheypresentwhenplottedintheGISwitha stipulationthataclustercontain30villagesminimum. 3.2Climatologicalandactualrainfallpatterns Our12-yearrainfallclimatologyagreeswellwithlonger-termrainfallpatternsandpreviousresearchintheregiondescribedearlier.Allcorrelationcoefcientscalculated betweenlatitudeandlongitudeandthevegrowingseasonmonthsaresignicantat a = 0.05.Normally,earlyseasonrainfallisgreatestinthesouthernandwesternportions ofthestudyregion(Table 1 ),whilerainfallistypicallyhighestinthenorthandeastin JanuaryandinthesouthandeastinFebruaryandMarch.Plotsofthe12-yearaverage rainfallwithinthefourgroupsproducedbythehierarchicalclusteranalysisconrmthis spatialpattern(Fig. 2 ).Rainfalloverthegrowingseasontypicallyassumesaninverted U''shape,withlowesttotalsinNovember(50100mm)andMarch(100200mm)and highestinJanuary(200300mm)formostvillages(Fig. 2 ),althoughsomevillages receivelessrainfallinFebruarythaninMarch.Asdiscussedpreviously,themigrationof theITCZintoandoutoftheregionalongwiththenorthwestsoutheastorientationofTTTs thatcontributetomoistureadvectionandrainfallproducethesespatialpatternsofrainfall. Pearson'scorrelationcoefcientscalculatedbetweenallstudymonths,monthlyclimatology,andlatitudeandlongitudedemonstratethatgenerally,thenormalseasonal patternofrainfalldidnotoccurduringourstudyperiod(Table 1 ).Forexample,thesouth andwestshouldreceivemorerainfallthanthenorthandeastduringtherst2monthsof thegrowingseason.However,thelackofrainfallearlyinSeasons1and3inthesouthand westwasdetrimentaltotherstcropsplanted(FAO 2003 2005 ).Correlationsbetween latitudeandtheseasonalrainfallanomalies(notshown)illustratethatthenorthgenerally receivedmorerainfallduringthestudyperiodthanthesouth,particularlyinSeasons1 Table1 Pearson'scorrelationcoefcientsformonthlyrainfallateachvillage( n = 419)andlatitude; longitude NovemberDecemberJanuaryFebruaryMarch Season10.564;0.3650.479;0.6790.196;0.5160.698;0.246 0.742; 0.718 Season2X; 0.4080.626;0.396 0.435; 0.6100.176;0.146 0.358; 0.703 Season30.745;0.7550.539;0.457 0.173; 0.2350.161;0.253 0.199;X Climatology 0.684; 0.793 0.441; 0.4770.099;0.344 0.253;0.331 0.138;0.334 Allcorrelationsshownaresignicantat a = 0.05 Positivecorrelationsindicatenorthandeast;negativecorrelationsindicatesouthandwest 36NatHazards(2013)66:3149123

PAGE 7

and3.AstrongnegativecorrelationwithlongitudeforSeason2(notshown)indicatesthat thewest,ratherthannorthorsouth,receivedmorerainfallduringthisseason. Althoughthespatialpatternsofrainfallvariedfromyeartoyear,generally,lessrainfall thannormaloccurredwithinthestudyregion,andingthatagreeswithpreviousresearch describingtheextendeddroughtinsouthernAfricaduringthistime(RouaultandRichard 2005 ;Matheretal. 2008 ).Therewere365instanceswherelessthan65%ofnormalrainfall occurredinconsecutivemonthsforavillage,andthesedryperiodsoccurredwithintherst 2monthsofSeasons1and2,andJanuaryMarchofSeason3.Addingthe5-monthrainfall totalforeachseasonandsubtractingfromthe12-yearaverage(Fig. 3 ac)demonstrates thatadecitofmorethan150mmofrainfalloccurredinallthreeseasonsfor12villages inManicaProvince,while98experiencedthisdecitintwoofthethreeseasons.The combinedeffectsofElNino(Seasons1and3),anegativeSIOD(Season1)andpositive IOD(Seasons1and2)(Table 2 )couldhelptoaccountforthegeneralizedpatternof drynessexperiencedacrossthethreegrowingseasons. Yetevenduringthisdryperiod,manyvillagesalsoreceivedmorerainfallthannormal, especiallyduringSeason1as39%ofvillagesreceivedmorethan150mmofrainfall abovetheseasonalaverage(Fig. 3 a).ThesendingscomportwiththoseofKadomura ( 2005 )whodescribedabnormallyhighrainfalleventsinthesouthernAfricanregionduring 2003.Generally,TTTstendtodevelopeastofMadagascarandfailtobringmoistureinto MozambiqueduringElNinoyears.However,ouranalysisofsatellitedatashowsthatTTTs stretchedacrossthenorthernportionofthecountryduringSeason1toprovideseveral Fig.2 NovemberMarchrainfallclimatologyforvillagesinthestudyregionstratiedintothefourgroups resultingfromtheclimatologicalhierarchicalclusteranalysis NatHazards(2013)66:314937123

PAGE 8

multi-dayrainfallevents,particularlyinlateJanuaryandMarch2003.Abovenormal rainfalltotalsinearlyJanuaryandMarchwerealsocausedbythepassageofCyclones DelnaandJaphet,whicharediscussedinmoredetaillaterinthepaper.Ageneralnorthto-southgradientinrainfalloccurred(Fig. 3 a),anditisnotsurprisingthattheFAO( 2003 ) reportedsuccessfulharvestsinthenorth,butmajorcropfailuresresultedfromdrought conditionsinthesouthernportionofourstudyareasothatfoodaidwasrequired. Fig.3 CumulativerainfallanomaliesNovemberMarchfor a season1, b season2,and c season3 38NatHazards(2013)66:3149123

PAGE 9

TherainfallpatternsofSeason2differgreatlyfromSeason1asevidencedbyan insignicantcorrelationcoefcientbetweenthesetwoseasons( 0.051).Season2isthe driestofthethreeseasonsaccordingtotheseason-averagedrainfallanomaliesforeach villageas41%ofvillageshadadecitofmorethan150mm(Fig. 3 b),andnearlyone quarterofthevillagesreceivedlessthan75%ofnormalrainfallacrosstheseason.Villages inNampulaandCaboDelgadoProvincesexperiencedthelowestpercentageofnormal rainfallacrosstheseason.Despitetherelativelydryseason,thecentralportionofthe countryreceivedmorerainfallthannormalinthebeginningand/orendoftheseason.In November,116villagesinTete,Zambezia,andNiassaProvincesreceivedmorethan150% ofnormalrainfall,whilethisexcessoccurredin37villageswithinTeteandManicaduring March(Fig. 4 ).Cropharvestsweremixedinthisregion,particularlyacrossZambezia,and farmersreplantedtwoandthreetimesduetotheirregularrain(FAO 2004 ). RainfallanomaliesaveragedoverSeason3exhibitstatisticallysignicantcorrelations withSeason1(0.443)andSeason2(-0.380).Earlyseasonrainfallwaswellabovenormal inthenorth,andrainfalldecreasedasSeason3progressedformostvillages.Februaryand March2005aretwoofthethreedriestmonthsinthestudy.Soilsinthenorthretainedthe early-seasonmoisturethroughoutthedriermonths,socropharvestsweregood(FAO 2005 ).However,verydryconditionsoccurredinTete,ManicaandZambeziaProvinces, whererainfalltotalsinconsecutivemonthswerelessthan50%ofnormal.Rainfall shortagesofmorethan150mmoccurredin31%ofvillagesinthestudyregion(Fig. 3 c). Onceagain,poorharvestsduetothedroughtmadefoodaidnecessaryforvillagesinthe southernportionofthestudyregion(FAO 2005 ). Table2 Teleconnectionindicesforeachstudymonth SOIaIODbSIODcNov2002 2 6.01.010 2 0.07 Dec2002 2 10.60.695 2 1.06 Jan2003 2 2.00.235 2 1.21 Feb2003 2 7.4 0.045 2 0.68 Mar2003 2 6.8 0.093 2 0.11 Nov2003 2 3.40.486 0.08 Dec20039.8 0.493 1.01 Jan2004 2 11.60.612 0.29 Feb20048.6 0.616 0.02 Mar20040.20.0710.55 Nov2004 2 9.3 2 0.0050.77 Dec2004 2 8.0 2 0.4960.10 Jan20051.8 2 0.9750.57 Feb2005 2 29.1 2 0.8560.47 Mar20050.2 0.755 2 0.36 BoldvaluesindicatethatdrynessshouldprevailinthestudyregionaccordingtopriorstudiesaSouthernOscillationIndexdataobtainedfrom http://www.cpc.noaa.gov/products/analysis_monitoring/ ensostuff/ensoyears.shtmlbIndianOceanDipoledataobtainedfrom http://www.jamstec.go.jp/frsgc/research/d1/iod/cSubtropicalIndianOceanDipoledataobtainedfrom http://www.jamstec.go.jp/res/ress/behera/iosdindex. html NatHazards(2013)66:314939123

PAGE 10

Thehierarchicalclusteranalysisbasedonthepercentageofnormalrainfallreceivedby villagesduringthe15studymonthsproducesfourclusters,inadditiontotheregions identiedasreceivingmorethan150mmofrainfallfromthetwocyclonesdiscussedin thenextsection.Thegeneralspatialpatternsofthefournon-cycloneclustersdemonstrate thatacrossthestudyperiod,rainfallpatternsaredistinctinthefarnortheastwithinCabo DelgadoProvince,thenorthcentralportionofthecountryinCaboDelgadoandNampula Provinces,thewesternportionofthecountryinthehigherelevationsneartheborderwith Malawi,andthesouthernreachesofthestudyareaincludingportionsofZambezia,Tete, Manica,andSofalaProvinces(Fig. 4 ).Graphsoftheaveragepercentageofnormalrainfall foreachgroupsupportourndingthatvillagesinthenorthreceivedmorerainfallduring thestudyperiodthanthoseincentralMozambique,andthatmorerainoccurredwithin Season1thantheothertwoseasons. 3.3Tropicalcycloneactivity AlthoughseveraltropicalcyclonespassedthroughtheMozambiqueChannelduringour studyperiod,onlyDelnaandJaphetmadelandfallinMozambique,whichisfewer landfallsthanwouldbeexpectedclimatologicallyoverthe3-yearstudyperiod.ElNino conditionstypicallyleadtofewerlandfallsinMozambique(Vitartetal. 2003 ),andthese conditionsoccurredduringSeasons1and3,sothetwolandfallsweanalyzebothoccurred Fig.4 Averagerainfallreceivedeachstudymonthforthefourgroupsproducedbythehierarchicalcluster analysisutilizingrainfallanomaliesandtwogroupsreceiving150mmormorerainfallfromatropical cyclone 40NatHazards(2013)66:3149123

PAGE 11

duringanElNinoyear.BothstormsformedintheMozambiqueChannelbuttookunusual tracksasneithermoveddirectlysouthwardandoutofthechannelasisthetypicaltrackof acycloneformingwithinthechannel(Reason 2007 ). TheTRMMdatashowthatby00UTCon31December,rainfallfromDelnabeganto affectthestudyregion.LandfalloccurredlaterthatdaynearAngocheinNampulaProvincewithmaximumsustainedwindsof28ms1(55kt).TheJTWC( 2003 )reportedand theTRMMdataconcurthattheremnantsofDelnamovedinlandwithaloopingtrack (Fig. 5 )sothatthesystemremainedoverthestudyregionforseveraldays,bringing oodingrainstovillagesinNampulaandZambeziaProvinces.AccordingtoBelletal. ( 2003 ),100,000peoplewerelefthomelessand34,000haofcropsweredestroyedin Mozambique.Ouranalysisrevealsthat104villagesinourstudyregionreceivedmorethan 150mmofrainfallovera7-dayperiod(Fig. 5 ).Insomeareas,morethan200mmof rainfalloccurredin1day,and29villagesreceivedmorethan300mmfromthissystem. Incontrast,Japhetaffectedasmallerportionofourstudyregionasittrackedstraight throughtheregionwithoutrecurving(Fig. 5 ).Landfalloccurredon2Marchsouthof VilanculosinInhambaneProvincewherewindswereestimatedtobe44ms1(85kt).The Fig.5 TracksofcyclonesDelnaandJaphetandthestorm-totalrainfallproducedbyeachcyclone NatHazards(2013)66:314941123

PAGE 12

JTWC( 2003 )andKadomura( 2005 )reportedthatstructuraldamageandoodingoccurred inareasoutsideofourstudyregion,andinseverallocationsinSofalaProvincenot sampledbythe2002and2005surveys.Conversely,rainfallinareasofSofalaandManica Provincesthatareincludedinourstudyregionhelpedtoalleviatedroughtconditions (JTWC 2003 ).Althoughwendthat30villagesreceivedmorethan150mmofrain,none receivedmorethan200mmonany1day.TheFAO( 2003 )reportedthat,forsomeareas, rainsfromJaphetwerefavorableforcropsandallowedforagoodharvest. 4Combinedrainfallandsocio-economicanalyses 4.1Socio-economicdata Mozambiqueprovidesagoodopportunitytostudysocio-economicimpactsofnatural hazardsinemergingeconomies.Asdescribedearlier,povertyandfoodinsecurityaffect themajorityofruralMozambicansandvulnerabilitytoextremeweathereventsishigh (Sheldon 2002 ;MADER 2005 ).Foodshortagesarecommon,andapproximately41%of ruralchildreninMozambiquewerestuntedduetomalnutritionin2003(UNICEF 2006 ). Thecentralandnorthernareasofthecountry,whicharetheregionsexaminedinthiscase study,experiencesomeofthehighestpovertyratesinthecountry. Economicglobalizationhasarguablyworsenedtheeconomicpositionofruralagriculturalists.InlinewithInternationalMonetaryFund(IMF)recommendations,Mozambiquebeganaseriesofstructuraladjustmentprogramsinthe1980sandthecountry maintainsaneoliberaleconomicagenda.Neoliberalismhashadseveralimpactsonthe agriculturalsector,includingtheliftingagriculturaltradebarriersandtheeliminationof governmentsubsidiestoruralproducers.Mozambiquehasearnedthepraiseofleading internationalnancialinstitutionsforitscommitmenttopovertyreductionthroughmarketbasedeconomicgrowth.However,evidencesuggestseconomicinsecurityandinequality inMozambiquedeepenedafterstructuraladjustmentduetoincreasingfoodprices,lackof agriculturalsubsidies,decliningemploymentopportunities,andthedismantlingofsocial protectionsforthepoor(Pfeifferetal. 2007 ;Silva 2007 2008 ;Pitcher 2002 ;Wuyts 2003 ). Mozambiqueisparticularlyvulnerabletodroughts,cyclones,andoods,thefrequency andseverityofwhicharepredictedtoincrease(IPCC 2007 ;EriksenandSilva 2009 ). Recentnaturaldisastershavedestroyednumerousroads,homesandschools.Forexample, oodinginFebruaryandMarchof2000,includingtheeffectsofCycloneEline,costthe Mozambicaneconomyanestimated$700million,including$428millioninreconstruction costs(WorldBank 2000 ).Ruralagriculturalistsborethebruntofthedisaster,suffering cropandpropertylossesthatgreatlylimitedtheirabilitytocopeinthedisaster'saftermath (ChristieandHanlon 2001 ).IncentralMozambiquealone,157villageswereooded (Steinbruchetal. 2002 ).Humanitarianassistanceisregularlyrequiredtofeedcommunities devastatedbysuchevents(Sheldon 2002 ). Tocalculatethechangeinincomeoverthestudyperiod,ouranalysisuseshouseholdlevelmicro-datafromthelongitudinalNationalAgriculturalSurveyofMozambique ( TrabalhodeInque ritoAgr cola, TIA).Thesurveywasconductedin2002and2005bythe MozambicanMinistryofAgricultureandRuralDevelopment(MADER 2002 2005 )and USAID.Thenationallyrepresentativehouseholdsamplewasdrawnusingastratied, clusteredsampledesign.Thesampleisstratiedbythecountry's10majorgeographical regionsandagro-ecologicalzones.Thecurrentstudyutilizeslongitudinaldatafor2,907 householdslocatedin419ruralvillagesacrosscentralandnorthernMozambique.The 42NatHazards(2013)66:3149123

PAGE 13

geographiccoordinatesofthesevillagesareobtainedfromtheTIAsurveysandthe locationsaremappedwithintheGISandcross-referencedusingthegeographicinformationfromthe1997Census.Thethree-yearspacingoflongitudinalhouseholdsurveys generallyprovidesanappropriateintervaltostudybehaviorchangesandagronomicshifts inhouseholdactivities(Carlettoetal. 2010 ). UsingtheTIA,eachhousehold'stotalincomeforeachyearisderivedbycalculating incomefromagriculturalcropandlivestockproduction,agriculturalwagelabor,nonagriculturalwagelabor,non-agriculturalself-employment,saleofnaturalresources,and pensionsandremittances.Sincethebulkoffoodeateninruralareasisproducedby householdsfortheirownconsumption,agriculturalproductionincomeiscalculatedasthe sumofcropandlivestocksalesandimputedincomefromownproductionconsumedby thehousehold.Agriculturalwagelaborconsistsofpart-timeandfull-timeoff-own-farm agriculturalemployment.Non-agriculturalwagelaborconsistsofpart-timeandfull-time off-farmemploymentinboththeformalandinformalsectors.Self-employmentincome consistsofnetbusinessprotsfrombothformalandinformalactivities.Naturalresource incomeconsistsofnetprotsearnedbyharvestingandsellingresource-basedgoods includingforestproducts,charcoal,huntedgame,andsh.Pensionandremittanceincome consistsofmoneyorgoodsreceivedfromgovernmentassistanceprogramsorfromfamily memberslivingandworkingawayfromhome.Theincometotalsarecalculatedfollowing theproceduresusedbyMatheretal.( 2008 )tomaketheresultscomparablewithprevious studiesusingtheTIAsurveys.Allincomeguresforhouseholdsarepopulationweighted. Percapitaincomeguresarethencalculatedbydividingahousehold'stotalincomebythe numberofitsmembers.Villagemeanpercapitaincomeisderivedbyaveragingtheper capitaincomeforhouseholdswithineachvillage.The2005incomeguresareination adjustedto2002levels.Thehouseholdchangeinpercapitaincomeiscalculatedasaratio of2005adjustedincomevaluesto2002values.Weremovealloutlierswithvaluesmore thanthreestandarddeviationsfromthemean(33householdsand5villages).Weusethe naturallogoftheincomechangevalueinourstatisticalanalysestonormalizethe distribution. Weemploytwotechniquestotestourhypothesisthatanassociationexistsbetween rainfallpatternsandchangesinincomeforourstudyregion.First,wecalculatePearson's correlationcoefcientsbetweenincomeandthepercentageofnormalrainfallforthe15 studymonthstoidentifywhichmonthshavethestrongestassociationswithchangein income.Second,weexplorespatialpatternsinthechangeinincomeusingtheresultsofthe secondclusteranalysisthatgroupedvillagesaccordingtothepercentageofnormalrainfall receivedacrossallstudymonths.Werstplacevillagesreceivingatleast150mmof rainfallfromcyclonesDelnaorJaphetintotheirowngroups.Theremainingvillagesare thenenteredintotheclusteranalysis.Afterverifyingthatthedatawithineachgroup assumeanormaldistribution,weperformaseriesofindependentsamples t testssothatthe averagechangeinincomeforeachofthesixgroupsisevaluatedagainsttheotherve. 4.2Resultsofcombinedanalysis Themajority(73%)ofvillagesinoursampleexperienceoveralldeclinesinincome (Fig. 6 ).Medianhouseholdincomefallsby63%overthe20022005studyperiod.Various factors,includingsocial,political,andhistoricalelements,inuencechangesinhousehold economicwell-being.However,aspreviouslynoted,thistimeperiodwasmarkedby numerousextremeweatherevents(bothwetanddry)inourstudyarea,makingitlikely thattheseeventscontributetothepooreconomicperformance.Thenortherncoastal NatHazards(2013)66:314943123

PAGE 14

provinceofNampulafaredtheworst,withthemedianhouseholdincomedecreasingby 71%.ThenorthernregionofMozambiqueischaracterizedbyhighpovertylevels(Bru ck andSchindler 2009 ),sodecliningincomesinthisregionsignifyextremehardshipfor households.ThesoutherncoastalprovinceofSofalaexperiencedthelowestdecreasesin income,butthemedianhouseholdincomestillfellby47%.Ingeneral,householdincome fellbysubstantialamountsinallsevencentralandnorthernprovincesincludedinthis studyandruralpovertyincreasedoverthetimeperiodunderinvestigation(Cunguaraand Hanlon 2010 ). Ourresultsshowthatcyclonescanhavedifferingeffectsonchangesinincome.Wend statisticallysignicantcorrelationsbetweenchangeinincomeandrainfallinDecember 2002( 0.197),January2003( 0.262),andfromDelna( 0.287).Highrainfalltotalsin December2002andJanuary2003areprimarilyassociatedwiththepassageofDelna, thoughtheTRMMdatashowthatTTTsdidcontributetomulti-dayrainfalleventsin northernMozambiquetotalingmorethan100mminlateJanuary2003.Themedian incomeforthe104villagesintheDelnaregiondecreasedby46%,whichisthehighest economiclossofallsixgroupsexaminedthrough t tests(Table 3 ).Furthermore,theresults ofthe t testsindicatethatthislossisstatisticallydifferentfromtheothervegroups (Table 4 ).Althoughwecannotsayforcertainthatnoothereventsinthisregionimpacted incomenegativelypriortothe2005survey,ourndingssuggestthatdamagecausedby Delnaimpactedtheregionmorethan2yearsafterthestorm'spassage. Fig.6 Theratioof2005incometo2002incomeforvillagesinthestudyregion 44NatHazards(2013)66:3149123

PAGE 15

PreviousresearchhasdocumentedthatrainfallproducedbycyclonescanhavelonglastingnegativeimpactsinMozambique,asin2000whenanestimatedtwomillionpeople wereaffectedbytheoodstowhichcycloneElinecontributed(Vitartetal. 2003 ).Interms ofthemaximumsustainedwindspeedatlandfall,DelnawasaweakercyclonethanEline, whosemaximumsustainedwindsatlandfallwerenearlytwicethatofDelnaat51ms1(100kt).Thus,themajorityofdamagesufferedbyvillagesintheDelnaregionwaslikely duetotheheavyrainfallandnotfastwinds.TheJTWCreportsconrmthatDelnacaused oodinginNampulaandZambeziaProvinces.Ourresultssuggestthatheavyrainfallfrom theremnantsofarelativelyweakcyclonecanhavelong-lastingdetrimentaleffectson economicwell-beinginthisregionofMozambique. Ontheotherhand,thepassageofJaphetmayhavebenettedvillagesinourstudy region.Thecorrelationcoefcientsshowthatvillagesfaringbettereconomicallyreceived morerainfromJaphet(0.144)andoverallinMarch2003(0.193).Mostofthe30villages intheJaphetregionhadadecreaseinincome.Yet,thislossonaverageisthesmallestof thesixregionalgroups(Table 3 ),andtheresultsofthe t testsindicatethatthismeanis signicantlydifferentfromthevillagesintheDelnaregionaswellasinthenorthcentral, south,andwest(Table 4 ).Wenotethatbenecialrainfallmaynotbetheonlyexplanation fortherelativeeconomicsuccessofvillagesintheJaphetregion,particularlycompared withthoseintheadjacentsouthernregionthatexperiencedsimilardryconditionsin Seasons2and3(Fig. 4 ).AChi-squaretestforindependenceindicatedthathouseholdsin theJaphetregionweremorelikelytohaveahouseholdmembergainsalariedemployment between2002and2005thanthoselocatedoutsidetheregion, X2(1, N = 2907) = 7.7, Table3 Statisticsforgroupsenteredintotheindependentsamples t tests Number ofvillages Meanratio of20052002 income(%) Naturallogof incomeratioaSDSEofthemean Japhet3096.4 0.0370.9800.179 Northeast3387.4 0.1350.6820.124 South12470.8 0.3460.7190.065 West6366.1 0.4140.8090.103 NorthCentral6563.4 0.4550.5230.066 Delna10446.2 0.7710.6420.063aNaturallogsofthemeanincomeratiovaluesforvillagesareutilizedintheanalysissothatthedata assumeanormaldistribution Table4 Signicancevaluesoftheindependentsamples t tests JaphetNortheastSouthWestNorthCentral Northeast0.655 South 0.054 0.140 West 0.0520.089 0.576 NorthCentral 0.0090.028 0.2380.737 Delna 0.0000.0000.0000.0010.002 Boldvaluesindicatethatthedifferenceinthemeansofthetwogroupsisstatisticallysignicantat a = 0.05, italicsindicatesignicanceat a = 0.10 NatHazards(2013)66:314945123

PAGE 16

p = 0.001.Thiscouldpossiblybeduetoemploymentgeneratedbyreconstructionprojects intheaftermathofthecyclone,similartothatwhichoccurredafterCycloneElinein2000 (ChristieandHanlon 2001 ).Interestingly,asimilartestshowednostatisticalrelationship betweenhouseholdlocationintheDelna-impactedareaandahouseholdmember acquiringsalariedemployment, X2(1, N = 2907) = 2.3, p = 0.13.Thedifferencesin salariedemploymentgainsafterthesetwocyclonessuggestthatfutureresearchshould explorethecombinedeffectsofweatherandsocio-economicfactorsonincomeinthe studyregion. Cyclonesarenottheonlymechanismbywhichabnormallyhighrainfalltotalscan occurinMozambique.InNovember2003,40villagesinTeteandZambeziaProvinces receivedmorethan200%oftheirnormalrainfall.Thisisthehighestpercentofrainfall receivedduringthestudyperiodthatwasnotduetothepassageofatropicalcyclone,and thenegativecorrelationcoefcientbetweenNovember2003rainfallandchangeinincome ( 0.128)indicatesthatvillagesaffectedbythisrainfallfaredpoorlyascomparedtoother villagesinthestudyregion.GiventhatTeteandZambeziaProvinceshavethehighest elevationsinthestudyregion,aplausibleexplanationforthestrongnegativeassociation betweenchangeinincomeandNovember2003rainfallisthatoodingoccurredthat damagedcrops.Insupportofourndings,FAO( 2004 )reportedthatTeteandZambezia hadlowagriculturalproductionoverallduringSeason2.Additionally,moremonths exhibitstatisticallysignicantnegativeratherthanpositivecorrelationswithincomeratios, andthecorrelationcoefcientsforthenegativevaluesarehigherthanthoseforthepositive values.Thus,wendthatevenduringthisdroughtperiod,receivingabove-normalrainfall, whetherfromtropicalcyclonesorotherconvectivesystems,tendstobeassociatedwith higherincomelossesthanreceivingbelow-normalrainfall. 5Conclusions Thisstudycomparesa12-yearrainfallclimatologywithactualrainfallpatternsduring threeagriculturalgrowingseasonsandrelatesthesepatternstoincomechangewithinrural Mozambicanvillages.Wendthatrainfallvariesregionallyoverthethreeseasons.During Season1,anorth-to-southgradientinrainfallisobservedandtwotropicalcyclonesaffect thestudyregiondifferentlywithDelnacausingoodinginthenorthandJaphetbringing benecialrainstothesouth.Seasons2and3aredrieroverall,withabove-normalrainfall totalsatthebeginningandendofSeason2inthewest,andrainfallwellabovenormaland decreasingtobelownormalinthenortheastandnorthcentralduringSeason3.Our ndingsabouttherelationshipbetweenrainfallanomaliesandsocio-economicwell-being goagainstsomegenerallyheldassumptionsintheliteratureonclimatemitigationand adaptation.First,cyclonesarenotalwaysdetrimental:Japhetmayhavehadsomepositive effectsontheeconomicpositionofhouseholds,eitherthroughbenecialrainor employmentcreatedbyreconstructionefforts.Second,gettingrainduringadroughtperiod isnotalwaysassociatedwithimprovedeconomicwell-being,especiallywhenrainfallis wellabovenormal.However,receivingmorerainfallattheendofthegrowingseason ratherthaninthebeginningormiddlewhendroughtconditionsexistmayhavepositive effectsonharvestsand,thus,theeconomicperformanceofhouseholds. Establishingtherelationshipbetweenrainfallanomaliesandsocio-economicchangein Mozambiquehasimportantimplicationsforunderstandingvulnerabilitytonaturalhazards inruralareascharacterizedbyextremepovertyandaheavyrelianceonrain-fedagriculture.However,otherfactorsinadditiontorainfallpatternsandextremeweatherevents, 46NatHazards(2013)66:3149123

PAGE 17

suchastheacquisitionofsalariedemployment,impactincomechangesforruralhouseholds.Inordertobetterinformanti-povertyprogramsandenhancethesustainabilityof rurallivelihoods,futureanalysisshouldconsiderthedynamicrelationshipsbetweenother driversofsocio-economicchangeaswellasclimatevariability.Assuch,thisstudytakes animportantrststeptowardevaluatingtheextenttowhichweatherimpactseconomic well-beingbydemonstratingthevalueofcombiningweatherandsocio-economicdatain ordertobetterunderstandpatternsofvulnerability.Acknowledgments TheauthorsthankDavidMather,DuncanBoughton,andEllenPayongayongfor providingaccesstothe2002and2005RuralHouseholdSurvey(TIA)forMozambiqueandgenerously sharingtheirexpertisewiththesedata.TheauthorsalsoacknowledgetheUnitedStatesAgencyforInternationalDevelopment(USAID)/MozambiquefortheirsupportoftheTIAdatacollectionandanalysisin Mozambique.Weappreciatethecommentsfromtwoanonymousreviewersthatwerehelpfulinreorganizingthemanuscriptintoitscurrentform.WealsothankStewartDuncanforfeedbackonanearlierdraftof thepaper.ReferencesAdeyewaZD,NakamuraK(2003)ValidationofTRMMradarrainfalldataovermajorclimaticregionsin Africa.JApplMet42:331347 AngelJR(2006)TropicalstormsreduceddroughtinIllinoisin2005.TransIllStateAcadSci99:111124 AshKD,MatyasCJ(2012)TheinuencesofENSOandtheSubtropicalIndianOceanDipoleontropical cyclonetrajectoriesintheSouthIndianOcean.IntJClimatol.doi: 10.1002/joc.2249 BeheraSK,YamagataT(2001)SubtropicalSSTdipoleeventsinthesouthernIndianOcean.GeophysRes Let28:327330 BellM,GroverE,HoppM,KestinT,LyonB,SethA(2003)ClimateImpacts.InternationalResearch InstituteforClimateandSociety. http://iri.columbia.edu/climate/cid/Feb2003/impacts.html .Accessed 9July2011 BrouwerR,NhassengoJ(2006)Aboutbridgesandbonds:communityresponsestothe2000oodsin MabalaneDistrict,Mozambique.Disasters30:234255 BrownME,FunkCC(2008)Foodsecurityunderclimatechange.Science319:580581.doi: 10.1126/ science.1154102 Bru ckT,SchindlerK(2009)Smalholderlandaccessinpost-warnorthernMozambique.WorldDev 37:13791389 CarlettoG,BeegleK,HimeleinK,KilicT,MurrayS,OseniM,ScottK,SteeleD(2010)Improvingthe availability,qualityandpolicy-relevanceofagriculturaldata:thelivingstandardsmeasurement studyintegratedsurveysonagriculture.WorldBank http://typo3.fao.org/leadmin/templates/ess/ pages/rural/wye_city_group/2010/May/WYE_2010.2.1_Carletto.pdf .Accessed13July2011 ChristieF,HanlonJ(2001)Mozambiqueandthegreatoodof2000.JamesCurrey,Oxford CookC,ReasonCJC,HewitsonB(2004)Wetanddryspellswithinparticularlywetanddrysummersinthe SouthAfricansummerrainfallregion.ClimRes26:1731 CunguaraB,HanlonJ(2010)PovertyisnotbeingreducedinMozambique.CrisisStatesWorkingPaperno. 47.CrisisStatesResearchCenter,LSE,London. http://eprints.lse.ac.uk/28467/ .Accessed7July2011 DinkuT,CeccatoP,Grover-KopecE,LemmaM,ConnorSJ,RopelewskiCF(2007)Validationofsatellite rainfallproductsoverEastAfrica'scomplextopography.IntJRemoteSens28:15031526.doi: 10.1080/01431160600954688 EriksenS,SilvaJA(2009)ThevulnerabilitycontextofasavannaareainMozambique:householddrought copingstrategiesandresponsestoeconomicchange.EnvironSciPolicy12:3552 FAO(2003)Specialreport2003:FAO/WFPcropandfoodsupplyassessmentmissiontoMozambique. FoodandAgricultureOrganizationoftheUnitedNations. http://www.fao.org/docrep/005/y9729e/ y9729e00.htm .Accessed9July2011 FAO(2004)Specialreport2004:FAO/WFPcropandfoodsupplyassessmentmissiontoMozambique. FoodandAgricultureOrganizationoftheUnitedNations. http://www.fao.org/docrep/007/J2679e/ J2679e00.HTM .Accessed9July2011 NatHazards(2013)66:314947123

PAGE 18

FAO(2005)Specialreport2005:FAO/WFPcropandfoodsupplyassessmentmissiontoMozambique. FoodandAgricultureOrganizationoftheUnitedNations. http://www.fao.org/docrep/008/J5510e/ J5510e00.htm .Accessed9July2011 FauchereauN,TrzaskaS,RouaultM,RichardY(2003)RainfallvariabilityandchangesinsouthernAfrica duringthe20thcenturyintheglobalwarmingcontext.NatHazards29:139154 Godfreys,TimoF,SmithM(2005)Relationshipbetweenrainfallandmicrobiologicalcontaminationof shallowgroundwaterinNorthernMozambique.WaterSA31:609614 HahnMB,RiedererAM,FosterSO(2009)Thelivelihoodvulnerabilityindex:apragmaticapproachto assessingrisksfromclimatevariabilityandchange-acasestudyinMozambique.GlobalEnviron Change19:7488 HallM,YoungT(1997)ConfrontingLeviathan:Mozambiquesinceindependence.OhioUniversityPress, Athens HanlonJ,SmartT(2008)DomorebicyclesequaldevelopmentinMozambique?.JamesCurrey,Oxford HansingoK,ReasonCJC(2009)ModellingtheatmosphericresponseoversouthernAfricatoSSTforcingin thesoutheasttropicalAtlanticandsouthwestsubtropicalIndianOceans.IntJClimatol29:10011012 HuffmanGJ,AdlerRF,BolvinDT,GuGJ,NelkinEJ,BowmanKP,HongY,StockerEF,WolffDB(2007) TheTRMMmultisatelliteprecipitationanalysis(TMPA):quasi-global,multiyear,combined-sensor precipitationestimatesatnescales.JHydromet8:3855.doi: 10.1175/jhm560.1 IPCC(2007)Climatechange2007:Thephysicalsciencebasis.ContributionofworkinggroupItothefourth assessmentreportoftheintergovernmentalpanelonclimatechange.CambridgeUniversityPress, Cambridge JointTyphoonWarningCenter(2003)2003annualtropicalcyclonereport.PearlHarbor,Hawaii JuryMR,PathackBMR(1993)Compositeclimaticpatternsassociatedwithextrememodesofsummer rainfalloversouthernAfrica19751984.TheorAppClimatol47:137145 JuryMR,WhiteWB,ReasonCJC(2004)ModellingthedominantclimatesignalsaroundsouthernAfrica. ClimDyn23:717726.doi: 10.1007/s00382-004-0468-z KadomuraH(2005)ClimateanomaliesandextremeeventsinAfricain2003.AfricanStudyMonographs, pp165181 KlinmanMG,ReasonCJC(2008)OnthepeculiarstormtrackofTCFavioduringthe20062007Southwest IndianOceantropicalcycloneseasonandrelationshipstoENSO.MetAtmosPhy100:233242.doi: 10.1007/s00703-008-0306-7 LeichenckoR,O'BrienK(2002)Thedynamicsofruralvulnerabilitytoglobalchange:thecaseofSouthern Africa.MitigAdaptStratGlobChange7:118 LobellDB,BurkeMB,TebaldiC,MastrandreaMD,FalconWP,NaylorRL(2008)Prioritizingclimate changeadaptationneedsforfoodsecurityin2030.Science319:607610.doi: 10.1126/science.1152339 LonfatM,RogersR,MarchokT,MarksFD(2007)Aparametricmodelforpredictinghurricanerainfall. MonWeatherRev135:30863097 MADER(Ministe riodaAgriculturaeDesenvolvimentoRural)(2002)Trabalhodeinque ritoagr cola2002 (Nationalagriculturesurvey2002).MADER,Maputo MADER(Ministe riodaAgriculturaeDesenvolvimentoRural)(2005)Trabalhodeinque ritoagr cola2005 (Nationalagriculturesurvey2005).MADER,Maputo ManatsaD,ChingombeW,MatsikwaH,MatariraCH(2008)ThesuperiorinuenceofDarwinsealevel pressureanomaliesoverENSOasasimpledroughtpredictorforsouthernAfrica.TheorAppClimatol 92:114.doi: 10.1007/s00704-007-0315-3 MatherD,CunguaraD,BoughtonD(2008)HouseholdincomeandassetsinruralMozambique,20022005: canpro-poorgrowthbesustained?DirectorateofEconomicsResearchWorkingPaperNo.66.Maputo, Mozambique:RepublicofMozambique. http://www.aec.msu.edu/fs2/mozambique/wps66.pdf Accessed7July2011 MatyasCJ(2010)Associationsbetweenthesizeofhurricaneraineldsatlandfallandtheirsurrounding environments.MeteorolAtmosPhys106:135148.doi: 10.1007/s00703-009-0056-1 MirzaM(2003)Climatechangeandextremeweatherevents:candevelopingcountriesadapt?ClimPolicy 3:233248 MittlemanJH(2000)Theglobalizationsyndrome:transformationandresistance.PrincetonUniversity Press,Princeton O'BrienK,VogelC(eds)(2003)Copingwithclimatevariability:theuseofseasonalforecastsinsouthern African.AshgatePublishingHouse,Burlington PfeifferJ,Gimbel-SherrK,AugustoOJ(2007)Theholyspiritinthehousehold:pentecostalism,gender,and neoliberalisminMozambique.AmAnthropol109:688700 PitcherMA(2002)TransformingMozambique:thepoliticsofprivatization,19752000.CambridgeUniversityPress,Cambridge 48NatHazards(2013)66:3149123

PAGE 19

ReasonCJC(2007)TropicalcycloneDera,theunusual2000/01tropicalcycloneseasonintheSouthWest IndianOceanandassociatedrainfallanomaliesoversouthernAfrica.MetAtmosPhy97:181188.doi: 10.1007/s00703-006-0251-2 ReasonCJC,JagadheeshaD(2005)AmodelinvestigationofrecentENSOimpactsoversouthernAfrica. MetAtmosPhy89:181205.doi: 10.1007/s00703-005-0128-9 ReasonCJC,KeibelA(2004)Tropicalcycloneelineanditsunusualpenetrationandimpactsoverthe southernAfricanmainland.WeaForecast19:789805 ReasonCJC,HachigontaS,PhaladiRF(2005)Interannualvariabilityinrainyseasoncharacteristicsoverthe LimpoporegionofsouthernAfrica.IntJClimatol25:18351853 RichardY,FauchereauN,PoccardI,RouaultM,TrzaskaS(2001)20thcenturydroughtsinsouthernAfrica: spatialandtemporalvariability,teleconnectionswithoceanicandatmosphericconditions.IntJClimatol21:873885 RouaultM,RichardY(2005)IntensityandspatialextentofdroughtsinsouthernAfrica.GeophysResLet 32.doi: 10.1029/2005GL022436 SchipperL,PellingM(2006)Disasterrisk,climatechangeandinternationaldevelopment:scopeforand challengestointegration.Disasters30:1938 SheldonKE(2002)Poundersofgrain:ahistoryofwomen,work,andpoliticsinMozambique.Heinemann, Portsmouth ShongweME,vanOldenborghGJ,vandenHurkB,deBoerB,CoelhoCAS,vanAalstMK(2009) ProjectedchangesinmeanandextremeprecipitationinAfricaunderglobalwarming.PartI:southern Africa.JClim22:38193837.doi: 10.1175/2009jcli2317.1 SilvaJA(2007)Tradeandincomeinequalityinalessdevelopedcountry:thecaseofMozambique.Econ Geogr83:111136 SilvaJA(2008)Amultilevelanalysisofagriculturaltradeandsocioeconomicinequalityinrural Mozambique.ProfGeogr60:174189 SitholeA,MurewiCTF(2009)ClimatevariabilityandchangeoversouthernAfrica:impactsandchallenges. AfrJEcol47:1720 SteinbruchF,GallM,Jose F(2002)RemotesensingandGISfordocumentationandevaluationofthesocioeconomicandenvironmentalimpactoftheoods2000inCentralMozambique.TheInternational ArchivesofthePhotogrammetry,RemoteSensingandSpatialInformationSciences. http://webra. cas.sc.edu/hvri/pubs/2002_RemoteSensingAndGISforDocumentationAndEvaluation.pdf .Accessed7 July2011 SuggAL(1968)Benecialaspectsofthetropicalcyclone.JApplMet7:3945 TadrossM,SuarezP,LotschA,HachigontaS,MdokaM,UnganaiL,LucioF,KamdonyoD,MuchindaM (2009)Growing-seasonrainfallandscenariosoffuturechangeinsoutheastAfrica:implicationsfor cultivatingmaize.ClimRes40:147161.doi: 10.3354/cr00821 ThomasDSG,TwymanC,OsbahrH,HewitsonB(2007)Adaptationtoclimatechangeandvariability: farmerresponsestointra-seasonalprecipitationtrendsinSouthAfrica.ClimChange83:301322.doi: 10.1007/s10584-006-9205-4 UNICEF(2006)ChildhoodpovertyinMozambique:asituationandtrendanalysis.UNICEF,Maputo UsmanMT,ReasonCJC(2004)DryspellfrequenciesandtheirvariabilityoversouthernAfrica.ClimRes 26:199211 VitartF,AndersonD,StockdaleT(2003)SeasonalforecastingoftropicalcyclonelandfalloverMozambique.JClim16:39323945 WashingtonR,PrestonA(2006)ExtremewetyearsoversouthernAfrica:roleofIndianOceanseasurface temperatures.JGeophysResAtmos111.doi: 10.1029/2005jd006724 WashingtonR,ToddM(1999)Tropical-temperatelinksinsouthernAfricanandSouthwestIndianOcean satellite-deriveddailyrainfall.IntJClimatol19:16011616 WattersonIG(2009)Componentsofprecipitationandtemperatureanomaliesandchangeassociatedwith modesoftheSouthernHemisphere.IntJClimatol29:809826.doi: 10.1002/joc.1772 WisnerB(1979)FloodpreventionandmitigationinthePeoplesRepublicofMozambique.Disasters 3:293306 WorldBank(2000)ApreliminaryassessmentofdamagefromtheoodandcycloneemergencyofFebruaryMarch2000.TheWorldBank,Washington WorldBank(2010)Africadevelopmentindicators2010.TheWorldBank,Washington WuytsM(2003)TheagrarianquestioninMozambique'stransitionandreconstruction.In:AddisonT(ed) FromconicttoreconstructioninAfrica.OxfordUniversityPress,London,pp141154 NatHazards(2013)66:314949123