DaresSalaamClimateProfile:FullTechnicalVersion

Preparedby:UniversityofCapeTown

November2017

ForenquiriesregardingthisClimateProfile,pleasecontactLisavanAardenne(lisa@csag.uct.ac.za)orLorenaPasquini(lorena.pasquini@gmail.com)

1

DaresSalaamclimateprofileHistoricclimateDar es Salaam is located on the Tanzanian coast and has a tropical climate with relatively hightemperatures,highhumidityandannual rainfallover1000mm. Rainfalloccursallyearroundbutprimarilyduringtworainyseasons:theshortrainyseasonfromOctobertoDecemberandthemain/longrainyseasonfromMarchtoMay.LessrainfalloccursduringJanuaryandFebruary,andthereisalongandrelativelydryseasonfromJunetoSeptember.ThewarmesttimeofyearisduringJanuaryandFebruary,andthecoolesttimeoftheyearoccursduringJulyandAugust;however,theseasoncycleisverysmall.MoredetailsareprovidedinFigure1below.

Figure1:1981to2010historicalaverageseasonalityforgridcelloverDaresSalaam.Meanmonthlytotal rainfall (mm/month) from the CHIRPS dataset depicted as blue bars, whiskers show +- 2standard deviations. Monthly mean daily maximum and minimum temperature from the WFDEIdatasetpresentedbytheredandgreenlinesrespectively.Dashedlinesrepresentthe+-2standarddeviationaroundthesemeans.

Theclimate isnot static, and rainfall and temperaturedisplayvariabilityonanumberofdifferenttimescales,fromdailytodecadal.Ontopofthis,theremayalsobeevidenceoflong-termtrendsintheclimate;however,itcansometimesbedifficulttodistinguishbetweenatrendandvariabilityatalongertimescale,unlessaverylonghistoricalrecordisused.

2RainfallRainfall varies on a number of time scales, from sub-daily to decadal. The diurnal cycle is animportantmode of variability for rainfall. Over this region, rainfall is predominantly convective innature(thunderstorms)andoccursduringthelateafternoon–earlyevening,butsomestormsmayalsodevelopovertheoceanatnight.Rainfall inDaresSalaamoccursduringallmonthsof theyear,however thebulkof the rain takesplaceduringtworainyseasons(Figure1).ThisbimodalrainfallpatternisduetothemigrationoftheInter-Tropical Convergence Zone (ITCZ) over the region. During January and February the ITCZ islocatedtothesouthoftheregionandtherainfallduringthisperiodaveragesaround110mm.TheITCZmoves north passing over the region duringMarch –May, resulting in the long rains duringwhereroughly610mmofrainfalls.Duringtheaustralwinter(June–September)theITCZissituatednorthoftheregionandlittlerainfallfalls(lessthan100mmduringthisperiod).TheITCZshiftsbacksouth passing over the region during October to November resulting in the short rains, whereroughly350mmofrainfallfallsonaverage.The average annual (July-June) total rainfall is around 1200mm, however the year-to-year orinterannualvariabilityis large(Figure2).Someyearsrecordaslittleas950mmwhileothersrecordmorethan180mm.Onmulti-yeartodecadaltimescalesDaresSalaamalsoexperiencesclearrainfallvariability (Figure 3). Much of this is related to large scale remote forcings such as the El NinoSouthernOscillation(ENSO)andtheIndianOceanDipole(IOD).OverthenortherncoastalareasofTanzania the positive phase (El Nino) of ENSO is generally associatedwith above-average rainfallduringtheshortrains(OND).Theoppositeistrueduringthecoolphase(LaNina)whichisgenerallyassociatedwithbelow-average rainfall. The increased rainfall associatedwithElNinoconditions isgenerallyduetoalonger-than-normalrainfallseasonandmorespecificallyanearlieronset.LaNinaconditions tend to be associated with later-than-normal onset of the rainy season1. The positivephaseoftheIODgenerallyleadstoincreasedrainfallduringtheshortrains,andthecoolerphasetodrier-than-normalconditions.Thelongrainsactindependentlytotheshortrains,anddonotexhibitanyconsistentrelationshiptoeitherENSOorIOD.Rainfall in Dar es Salaam has changed over time (Figure 4). The annual (July to June) rainfall hasdecreasedinmagnitudeby-26mm/decade,andthefrequencyofwetdayshasdecreasedbyover-6.5 days per decade. However, the rainfall within the two primary rainy seasons has changedindependently of each other. The long rains (March to May) (Figure 6) show an increase inmagnitudeorseasonaltotalrainfallamount,andthisislinkedtoastatisticallysignificantincreaseinthe daily intensity which has increased by nearly 6mm/day per decade. In contrast to the dailyintensity,thefrequencyofraindayshasseenadecreasingtrendthroughtimeofalmost4daysperdecade.Anegativetrendisalsoseenintheaveragedurationofwetspells,whichhavedecreasedbyalmost 0.3 days per decade. This decrease inwetspell duration ismirrored by an increase in theaveragedryspellduration,withsomeyearsshowingverylongperiodsbetweenrainevents.Therate

1 Kijazi, A. & Reason, C. (2005) Relationships between intraseasonal rainfall variability of coastal Tanzania and ENSO.TheoreticalandAppliedClimatology82:153-176.

3ofchangewasgenerallyhighestduringthe1980sand1990sformostoftheserainfallstatistics,afterwhichtherateofchangegenerallydecreased.Theshortrains(OctobertoDecember)(Figure7)showadifferentpicture.Thereisanegativetrendintheseasonaltotalrainfallof-15mm/decade.Thestrongestdecreasesoccurredduringthe1980sand1990safterwhichtherehasbeenaslightrecoveryinrainfalltotals.Thefrequencyofraindaysshowsasimilarpattern,withanegativetrendoverall,ofjustunder2daysperdecade.Theaveragewetspelldurationhasdecreasedby0.1daysperdecadewhich ismirroredbytheaveragedryspelldurationwhichhasincreasedbyalmost11daysperdecade.Heavyrainfallfrequencyalsoshowsanegative trend of almost 1 day per decade. All these statistics show that the strongest negativechangeforbothrainyseasonsoccurredduringthe1980sand1990s,andthattherehasbeenaslightrecoveryfromaround2000tothepresent.Theonestatisticnotinlinewiththesegeneraltrendsisthatofdailyaveragerainfall intensity,whichshowsanincreasethroughtimeandespeciallyduringthefirsttwodecades(1980sand1990s).AsummaryoftheseresultscanbefoundinTable1below)TemperatureTemperaturealsoshowsclearvariabilityonarangeoftimescales.Thediurnaltemperaturerange–or difference between themaximum andminimum temperaturewithin a 24-hour period – is animportantmodeof variability forDar es Salaam.AtDar es Salaam it averages just over 9 °C, butvariesbetweenseasons;thelargestdifferencesoccurinthesecondhalfofthedryseason,especiallySeptember(11.3°C),andthesmallestdifferencesoccurinApril(7.9°C),whichisthewettestmonthoftheyear(Figure1).Theseasonalcyclesofdailymaximumandminimumtemperaturearerelativelysmall,averaging3.5°Cand5 °C respectively (Figure1).Thewarmest temperaturesoccurduring JanuaryandFebruary(~32.5°Cmaximumand24°Cminimum)andthecoolesttemperaturesoccurduringthewinterdryseason(~29.5°and19°C).ThissmallseasoncycleisduetoDaresSalaambeinglocatedveryclosetotheequatorandonthecoastofawarmoceanwhichmodulatesthetemperaturethroughtheyear.Temperaturesshowlittleinterannualvariability,withannualaveragedailymaximumandminimumtemperaturesgenerallyvaryingbylessthanhalfadegreefromyeartoyear(Figure2).Thevariabilityforaspecificmonthorseasonisgenerallylargerthanthatforannualaveragetemperature,varyingbyasmuchas1°Cfromthelongtermmean.ThisinterannualvariabilityisoftenlinkedtotheENSOsignal, with warmer temperatures generally being associated with El Nino conditions, and coolertemperaturesbeingassociatedwithLaNinaconditions.Seefigure3forfurtherdetails.Focussingonthelonger-termmulti-decadaltimescale,DaresSalaamshowsaslightpositivetrendindailymaximumtemperaturethroughouttheyear,butthetrendisonlystatisticallysignificantduringthecoolerdryseason(+0.1°Cperdecade)(Figure8).Extremehotdays(dayswherethetemperatureexceeds 34.6 °C) occur historically from October through to April of the following year. Thesedaytimehoteventshaveseenaslightdecreaseinfrequency(ofnumberofhotevents)andduration(ofnumberofconsecutivehotdays),especiallyduringthe1980s(Figure9).

4There is a strongerwarming trend for averagenight-time temperatures from1979-2014, and thiswarming trend is statistically significant in all seasons except during the long rains (Figure 10).Extremewarmnights(i.e.nightswheretheminimumtemperatureis>24.5°)historicallyoccurfromDecembertoMarch,andtherehasbeenaveryclearincreaseinthefrequency(+11daysperdecade)andduration(ofnumberofconsecutivehotnights;+0.2daysperdecade)(Figure11).“Extremecold”days(whichforDaresSalaammeansdayswherethemaximumtemperatureis<28°C)historicallyoccurduring June toAugust,butoccasionallyoccurduringother timesof theyear.Thefrequencyofthesecolddayshasdecreasedoverthelast30yearsbyalmost8daysperdecadeandthedurationof“extremecold”spells(i.e.back-to-back“extremecold”days)hasdecreased,nowaveraging just a single day per year (Figure 12). “Extreme cold” nights (which for Dar es Salaammeansnightswheretheminimumtemperatureis<18.8°C)historicallyoccurprimarilyduringJulytoSeptember, and the frequency and duration of “extreme cold” night-time spells have alsoexperiencedastatisticallysignificantdecreaseof-7.6and-0.2daysperdecaderespectively,withthesharpestdecreaseoccurringduringthe1990s(Figure13).

5Table1:Summaryoftrendsintemperatureandrainfallattributes,DaresSalaam.TheTheil-Sen lineartrendslope(perdecade) isprovidedalongwiththeclimatologicalmeanvalue(inbrackets).TherainfallisfromtheDaresSalaamweatherstationandthetemperaturetrendsarecalculatedfromthegridcellvalueoverDaresSalaamfromtheWFDEIdataset.Valuesinbold indicateastatisticallysignificantchange.Red(blue)colour indicatesawarming(cooling)temperaturetrend.Green(brown)colourindicatesawetting(drying)precipitationtrend.

Temperature Jan–Feb Mar–May Jun–Sep Oct–Dec

Tmax[°C/decade] +0.05(mean=32.8)

+0.1(mean=31.1)

+0.1(mean=29.8)

+0.1(mean=31.9)

Tmin[°C/decade] +0.3(mean=24.1)

(+0.1)(mean=22.7)

+0.2(mean=19.5)

+0.3(mean=22.3)

Annual(July–June)

Tmaxextremehotevents[days] Frequency:-3.3Duration:-0.1Threshold:34.6°C

Tminextremehotevents[days]

Frequency:+11Duration:+0.2Threshold:24.5°C

Tmaxextremecoldevents[days]

Frequency:-7.9Duration:-0.1Threshold:28.0°C

Tminextremecoldevents[days]

Frequency:-7.6Duration:-0.2Threshold:18.8°C

Rainfall Jan–Feb Mar-May Oct-Dec

Totalrainfall[mm/decade] -1.7(mean=105.4) +24.2(mean=573.5) -15.6(mean=275)

Rainintensity[mm/day] +0.8(mean=11.9) +5.8(mean=36.3) +4.8(mean=37.7)

Raindayfrequency[days/decade]

-0.2(mean=9.7) -3.8(mean=23.6) -1.9(mean=10.6)

Heavyraindayfrequency[days/decade]

notrend(mean=4.3)

notrend(mean=16.5) -1.0(mean=7.8)

Wetspell[consecutivedays]

notrend(mean=1.5) -0.3(mean=2.0) -0.1(mean=1.2)

Dryspell[consecutivedays] -0.3(mean=14.2) +2.8(mean=22.9) +10.9(mean=53.9)

6

Climatechangeprojections

GlobalclimatemodelsProjectionsoffutureclimate,basedon15CMIP5GlobalClimateModels(GCMs)simulations2undertheRCP8.5pathway3,showaclearandstatisticallysignificant increase inbothdailymaximumandminimumtemperatureintothefuture(figure14&15).By2040temperaturesmaybebetween0.5-1°C(formaximumtemperatures)and0.75-1.75°C(forminimumtemperatures)warmerthanthecurrent climate, depending on the GCM. Note that two models simulate an unrealistic decadalvariability indailymaximumtemperature,butdonot change theoverallmessage (please refer toFigure14).Thenumberofextremehotdaysisprojectedtoincreaselinearlythroughtime,butthereis largedisagreementontherateofchange (Figure16).Somemodelsproject justasmallchange,whileothersproject thatup tohalf thedaysper yearwill exceed thehistorical 90thpercentileby2040.Thenumberofextremehotnights isalsoprojected to increase into the futureand there ismore agreement on the rate of change compared to dailymaximum temperature (Figure 17). By2040thefrequencyofthesehotnightsisprojectedtooccuronbetween100and200daysperyeardependingontheGCM.Rainfall is projected to remainwithin the historic range of variability until the second half of thecentury,whensomemodelsprojectan increase,whileacoupleprojectadecrease in rainfall;halfthemodelsprojectthatitwillremainwithinthecurrentrangeofnaturalvariabilityuptotheendofthecentury(Figure18).Themodelsprojectnochangeoraslight increaseindailyrainfall intensitytoward the end of the century (Figure 19). The frequency of heavy and extreme rainfall days isprojectedtoremainwithintherangeofnaturalvariabilityforthefirsthalfofthecentury(Figure20&21).Bytheendofthecenturysomemodelsprojectanincreasewhileafewprojectadecreaseinfrequency.Table2belowprovidesasummaryoftheseprojectedclimatechangesforDaresSalaam.

2ThefifthiterationoftheCoupleModelInter-comparisonProject(CMIP)isacoordinatedactivityamongstinternationalmodellingcentrestoproduceasuiteofclimatesimulationsusingcommonexperimentalparameters.CMIP5iscurrentlytheprimary sourceofglobal to regional scale climateprojectionsandextensively informed the IPCCFifthAssessmentReport(AR5)3Althoughthisemissions/developmentpathwayrepresentsthe“worst-casescenario”amongstthepathwayssimulatedby the IPCC CMIP5models, at this stage it is themost realistic reflection of the recent progression of anthropogenicemissions. It is presented here, in spite of the Paris Agreement, as effects of the latter’s commitments remain to beshown.

7

Table2:SummaryofGCMprojectedclimatechangesforkeyclimatevariables,DaresSalaamStatistic Annual

AverageTmax[°C]Increasing+0.5°Cto+1°Cby2040s,andbybetween2-4°Cbytheendofthecentury,butchangesevidentinnextdecade(note:twomodelsshowunexpecteddecadalvariability)

AverageTmin[°C] Increasing:+0.75°Cto+1.75°Cby2040s,andbybetween2–4.5°Cbytheendofthecentury,butchangesevidentinnextdecade

Tmaxextremehotevents[daysover themodel’shistorical 90thpercentile]

Increasing:linearincreaseofnumberofextremehotdaysintothefuture,withverylargespreadintherateofincreasedependingonthemodel.Thefrequencyby2040rangesbetweensmallchange(asmallincreaseinthenumberofextremehotdaysperyear),tohalf the days per year exceeding the model’s historic 90thpercentile.Frequencyby theendof thecentury rangesbetweenlessthandoubling,toalldaysperyearexceedingthethreshold.

Tmin extreme hot events[nights over the model’shistorical90thpercentile]

Increasing: steady rate of increase in number of extreme hotnights from present to the 2070s, afterwhich there is a slightlyslower increase.Relatively small intermodel spreadcompared todailyextremehotevents.By2040thefrequencyofextremelyhotnights (i.e. nights over the model’s historic 90th percentile)increasestobetween100and200daysperyear,andbytheendofthecenturyitvariesbetween220toalldaysperyear.

Rainfalltotals[mm]Normaltoincreasingordecreasingrainfall,by2040mostmodelsshow no change, but by the end of the century the intermodelspreadrangesfromwettingtodryingofaround200mm/year.

Rainintensity[mm/day]Normal to slightly increasing rainfall intensity, by 2040 mostmodels show no change, but by the end of the century a fewmodelsshowaslightincreaseinrainfallintensity.

Heavy rainfall frequency [daysofrainfall>10mm]

Normal to increasing or slightly decreasing heavy rainfallfrequency, by 2040 most models show no change or a slightincrease, but by the end of the century more models show anincreaseandafewshowadecreaseinfrequency.

Extreme rainfall frequency[days of rainfall > the model’shistorical90thpercentile]

Normal to increasing or slightly decreasing extreme rainfallfrequency, by 2040 most models show no change or a slightincrease, but by the end of the century more models show anincreaseandafewshowadecreaseinfrequency.

8StatisticallydownscaledprojectionsProjectionsoffutureclimate,basedon11statisticallydownscaledCMIP5GCMsimulationsundertheRCP8.5pathway,showaclearandstatisticallysignificant increase inbothminimumandmaximumtemperatureintothefutureandthechangesshouldalreadybestartingtobecomeevident(Figure22&23).By2040thedailymaximumandminimumtemperaturesmaybebetween1–1.75°Cand1–1.5 °Cwarmer than the current climate, respectively, andup to2.75–4.5 °Cand2.5–4.25 °Cwarmer by the endof the century, dependingon themodel selected. The frequencyof days andnights exceeding the historical extreme temperature threshold (90th percentile) is projected toincrease (Figure 24 & 25), however the extreme daily minimum temperatures are expected toincreasemorerapidlythantheextremedailymaximumtemperatures,especiallybeforethe2060s.Rainfallisprojectedtoremainwithinthehistoricrangeofvariabilityordecreaseoverthefirsthalfofthe21stcentury;bytheendofthecentury,allbuttwomodelsprojectadryingtrend(Figure26).Thisisalsoseen in themagnitude, intensityand frequencyof rainfallandextremerainfall (Figure27–29).Table3belowprovidesasummaryoftheseprojectedclimatechanges.Table 3: Summary of statistically downscaled GCMs projected climate changes for key climatevariables,DaresSalaamStatistic Annual

AverageTmax[°C]

Increasing+1°Cto+1.75°Cby2040s,andbybetween2.75–4.5°Cbythe end of the century, but changes may already be evident (note:modelsseparateintotwogroupswith6moreconservativeand5moreextremeprojections

AverageTmin[°C]

Increasing+1°Cto+1.5°Cby2040s,andbybetween2.5–4.25°Cbythe end of the century, but changes may already be evident (note:modelsseparateintotwogroupswith6moreconservativeand5moreextremeprojections.

Tmaxextremehotevents[daysover themodel’shistorical 90thpercentile

Increasing:gradual increase in frequencyupuntil2060,bywhich timethe frequency increasesby50%to100%.Thenmore rapid increase totheendofthecentury,wherethefrequencyrangesbetween120to225daysperyear.

Tmin extreme hot events[nights over the model’shistorical90thpercentile]

Increasing: steady rate of increase from present to the end of thecentury,frequencydoublestotriplesby2040,andreachesbetween150and275daysperyearbytheendofthecentury.

Rainfalltotals[mm]Normal todecreasing rainfall, by 2040mostmodels showno change,but by the end of the century all but two models project significantdrying.

Rainintensity[mm/day]Normaltodecreasingrainfall intensity,by2040mostmodelsshownochange, but by the end of the century all but two models projectsignificantdecreaseinrainfallintensity.

Rainfallfrequency[days]Normaltodecreasingrainfallfrequency,by2040mostmodelsshownochange, but by the end of the century all but three models projectsignificantdecreaseinrainfallfrequency.

Extreme rainfall frequency[days of rainfall > the model’shistorical90thpercentile]

Normaltodecreasingrainfallfrequency,by2040mostmodelsshownochange, but by the end of the century all but three models projectsignificantdecreaseinextremerainfallfrequency.

9SupportingevidenceThe above summary information is supported by rigorous analysis of observed and modelprojectionsdata.Moredetailsofthisanalysisandsupportingfigurescanbefoundbelow.DataThis study focusesonhowtheclimate forDaresSalaamhaschanged in thepastandhow itmaychange in the future due to anthropogenic climate change. Ideally one would like to base thehistoricalanalysisondatafromanumberofweatherstationstoobtainadetailedunderstandingofthelocalclimatesinthedifferentpartsofthecity.Unfortunatelytheonlypublicly-availableweatherstationdata forDaresSalaamareof insufficient lengthandquality touse in thisanalysis. Insteadthisanalysisreliesontemperaturedatafromagriddedproductcall theWATCHForcingDataERA-Interim (WFDEI)4 where the WATCH Forcing Data methodology is applied to ERA-Interim data(Weedonetal.2014)5.Itprovidesdataforthegloballandsurfaceat0.5°x0.5°coveringtheperiod1979-2014.ThedailyrainfalldatausedinthehistoricalanalysisisobtainedfromtheClimateHazardsGroup InfraRed Precipitationwith Station data (CHIRPS)6 (Funk et al. 2015)7. CHIRPS incorporates0.05°resolutionsatelliteimagerywithstationdatatocreateagriddedrainfalltimeseriesformostoftheglobe.Theversion2.0isusedinthisanalysiswhichprovidesdataona0.05°grid.Two different sets of climate change data are used to explore the possible future changes in theclimate due to anthropogenic climate change. The first set is an ensemble of 15 Global ClimateModels(GCMs)fromtheClimateModelIntercomparisonProjectionversion5(CMIP5)(alistofthemodelsandmodellinggroups isprovided in table4below).Daily rainfall,maximumandminimumtemperaturefromthehistoricalexperiment(1960-2005)andtheRCP8.5futureemissionexperiment(2006-2100)wereusedtoexplorehowthesevariablesareprojectedtochangeintothefuture.Thesecond set of climate change data is an ensemble of 11 statistically downscaled CMIP5 GCMs.Circulation fields from theGCMswere used as predictor variables,while theWFDEI daily rainfall,maximum and minimum temperature data were used as predictant datasets in a statisticaldownscalingmethodologycalledSelf-OrganisingMapbasedDownscaling(SOMD)developedbytheClimate SystemAnalysisGroup (CSAG) (Hewitson&Crane 20068). The downscaling provides dailyrainfall, maximum and minimum temperature for each GCM for the historical (1960-2005) andRCP8.5future(2006-2100)experimentata0.5°resolution.

4EUWATCH–DataforResearchers:http://www.eu-watch.org/data_availability5Weedon,G.P.,Balsamo,G.,Bellouin,N.,Gomes,S.,Best,M.J.&Viterbo,P. (2014)TheWFDEImeteorological forcingdata set: WATCH Forcing Data methodology applied to ERA-Interim reanalysis data. Water Resources Research, 50:7505–7514.6CHG–Data–CHIRPS:http://chg.geog.ucsb.edu/data/chirps/7Funk,C.,Peterson,P.,Landsfeld,M.,Pedreros,D.,Verdin,J.,Shukla,S.,Husak,G.,Rowland,J.,Harrison,L.,Hoell,A.&Michaelsen, J. (2015) The climate hazards infrared precipitation with stations – a new environmental record formonitoringextremes.ScientificData2,Articlenumber:150066.8Hewitson,B.C.&Crane,R.G.(2006)ConsensusbetweenGCMclimatechangeprojectionswithempiricaldownscaling:precipitationdownscalingoverSouthAfrica.InternationalJournalofClimatology26:1315-1337.

10A time series for the gridcell covering Dar es Salaam was extracted from each of the observeddatasetsandalsofromalloftheGCMandstatisticallydownscaleddata.Thesedatawereusedinalltheanalyses.

Table4:CMIP5modellingcentresandmodelsusedintheanalysis.Models in italicsarealsousedinthestatisticaldownscalingMODELINGCENTRE(ORGROUP) INSTITUTEID MODELNAME

Beijing Climate Center, China MeteorologicalAdministration

BCC BCC-CSM1.1

College of Global Change and Earth System Science,BeijingNormalUniversity

GCESS

BNU-ESM

CanadianCentreforClimateModellingandAnalysis CCCMA

CanESM2

CentreNationaldeRecherchesMeteorologiques/CentreEuropeen de Recherche et Formation Avancees enCalculScientifique

CNRM-CERFACS

CNRM-CM5

LASG, InstituteofAtmosphericPhysics,ChineseAcademyofSciences

LASG-IAP

FGOALS-s2

NOAAGeophysicalFluidDynamicsLaboratory

NOAAGFDL

GFDL-ESM2GGFDL-ESM2M

InstitutPierre-SimonLaplace IPSL IPSL-CM5A-MRIPSL-CM5B-LR

InstituteforNumericalMathematics INM INM-CM4

AtmosphereandOceanResearchInstitute(TheUniversityof Tokyo), National Institute for Environmental Studies,and Japan Agency for Marine-Earth Science andTechnology

MIROC

MIROC5

Japan Agency for Marine-Earth Science and Technology,AtmosphereandOceanResearchInstitute(TheUniversityof Tokyo), and National Institute for EnvironmentalStudies

MIROC

MIROC-ESMMIROC-ESM-CHEM

MaxPlanckInstituteforMeteorology(MPI-M) MPI_M MPI-ESM-LR

MeteorologicalResearchInstitute

MRI

MRI-CGCM3

11HistoricaltrendsandvariabilityanalysisThe analysis of historical trends and variability of key climate variables is presented below. Thisanalysis uses daily maximum and minimum temperature data obtained from the WATCH whichcovers the period 1979 - 2014. The rainfall dataset used is the CHIRPS dataset covers the periodJanuary1981–December2016.Thesegriddeddatasetswereusedsincethequalityand lengthofthe weather station record for Dar es Salaam was too poor to be used in this analysis. Derivedstatisticswerecalculatedattheseasonalandannualtimescale.ThesewereusedtoexplorethelongtermtrendsandvariabilityoftheclimateatDaresSalaam.

12

Figure2:TimeseriesofmonthlymeanmaximumandminimumtemperatureandtotalrainfallforthegridcelloverDaresSalaam.Redandgreencolouredlinesrepresenta12-monthrunningaveragefor maximum and minimum temperature respectively from the WFDEI dataset. Light blue barspresenttheannual(July–June)totalrainfallfromtheCHIRPSdataset.

13

Figure3:AssociationbetweenENSOandtheclimateatDaresSalaamthroughtime.Timeseriesofthe NINO 3.4 SSTmonthly anomalies is presented as the grey line; positive (El Nino) phases arecoloured red,while negative (LaNina) phases are shaded in blue. The black line in the top panelshowsthemonthlymeanmaximumtemperatureanomaliessmoothedwitha12-valuerunningmeanfromtheWFDEIdataset.Thesecondpanelshowsthesameasabove,butforminimumtemperature.TheblackbarsinthebottomtwopanelsshowtheMarch-MayandOctober–Decemberseasonaltotalrainfallanomalies(mm/season)fromtheCHIRPSdataset.

14

Figure4:Timeseriesandtrendindailyrainfall(pr>0.2mm)forthegridcelloverDaresSalaamfromtheCHIRPSdataset.Thetoppanelshowsthetiminganddurationofraineventsineachyear(July-June)oftherecord.Themiddlepanelshowstheannual(July-June)totalrainfall inbluebarsalongwith the Theil-Sen trend (red line) and the Lowess smoothmean (black line) and 95th confidenceinterval (dashed line). The bottompanel shows the same as above but for the annual (July-June)frequencyofraindays.

15

Figure5:TimeseriesandtrendinJanuary-FebruaryseasonalrainfallstatisticsforthegridcelloverDaresSalaamfromtheCHIRPSdataset.EachpanelshowsthatstatisticasabluebarsalongwiththeTheil-Sen trend (red line) and the Lowess smoothmean (black line) and 95th confidence interval(dashedline).

16

Figure6:TimeseriesandtrendinMarch-MayseasonalrainfallstatisticsforthegridcelloverDaresSalaamfromtheCHIRPSdataset.EachpanelshowsthatstatisticasabluebarsalongwiththeTheil-Sentrend(redline)andtheLowesssmoothmean(blackline)and95thconfidenceinterval(dashedline).

17

Figure7:TimeseriesandtrendinMarch-MayseasonalrainfallstatisticsforthegridcelloverDaresSalaamfromtheCHIRPSdataset.Eachpanelshowsthatstatisticasabluebarsalongwith theTheil-Sen trend (red line) and the Lowess smoothmean (black line) and 95th confidence interval(dashedline).

18

Figure8:Timeseriesand trend in seasonalaveragemaximumtemperature for thegridcelloverDaresSalaamfromtheWFDEIdataset.Timeseriesofseasonalmeanmaximumtemperature(bluedots);Theil-Sentrend(redline),theLowesssmooth(blackline),and95thconfidenceinterval(dashedlines).

19

Figure9:Timeseriesandtrendinthefrequencyofhotdays(Tmax>34.5)forthegridcelloverDaresSalaamfromtheWFDEIdataset.Thetoppanelshowsthetiminganddurationoftheseeventsineachyear(July-June)oftherecord.Themiddlepanelshowstheheatspellfrequencyornumberofdaysperyear(July–June)intheredbarsalongwiththeTheil-Sentrend(redline)andtheLowesssmooth(blackline)and95thconfidenceinterval(dashedlines).Thebottompanelshowsthesameasabovebutfortheaveragespelldurationornumberofconsecutivedaysperyear(July-June).

20

Figure10:TimeseriesandtrendinseasonalaverageminimumtemperatureforthegridcelloverDaresSalaamfromtheWFDEIdataset.Timeseriesofseasonalmeanminimumtemperature(bluedots);Theil-Sentrend(redline),theLowesssmooth(blackline)and95thconfidenceinterval(dashedlines).

21

Figure11:Timeseriesandtrendinthefrequencyofhotnights(tmin>24.5)forthegridcelloverDar es Salaam from theWFDEI dataset. The top panel shows the timing and duration of theseevents ineachyear (July-June)of the record.Themiddlepanel shows theheat spell frequencyornumberofdaysperyear(July–June)intheredbarsalongwiththeTheil-Sentrend(redline)andtheLowesssmooth(blackline)and95thconfidenceinterval(dashedlines).Thebottompanelshowsthesameasabovebutfortheaveragespelldurationornumberofconsecutivedaysperyear(July-June).

22

Figure12:Timeseriesandtrendinthefrequencyofcooldays(Tmax<28)forthegridcelloverDaresSalaamfromtheWFDEIdataset.Thetoppanelshowsthetiminganddurationoftheseeventsineachyear(July-June)oftherecord.Themiddlepanelshowstheheatspellfrequencyornumberofdaysperyear (Jan–Dec) in theredbarsalongwith theTheil-Sentrend (red line)andtheLowesssmooth(blackline)and95thconfidenceinterval(dashedlines).Thebottompanelshowsthesameasabovebutfortheaveragespelldurationornumberofconsecutivedaysperyear(July-June).

23

Figure13:Timeseriesandtrendinthefrequencyofcoldnights(Tmin<18.8)forthegridcelloverDar es Salaam from theWFDEI dataset. The top panel shows the timing and duration of theseevents ineachyear (July-June)of the record.Themiddlepanel shows theheat spell frequencyornumberofdaysperyear(Jan–Dec)intheredbarsalongwiththeTheil-Sentrend(redline)andtheLowess smooth mean (black line) and 95th confidence interval (dashed lines). The bottom panelshowsthesameasabovebutfortheaveragespelldurationornumberofconsecutivedaysperyear(July-June).

24

ClimateprojectionsvisualizationsGlobalClimateModelsThe plots below (Figures 14 to 29) are called plume plots and they are used to represent thedifferentlong-termprojectionsacrossthemultipleclimatemodelsintheCMIP5modelarchiveusedto inform the IPCCAR5 report. The plots showprojected variations in different variables for thegridcell over Dar es Salaam produced by an ensemble of 15 models. The blue colours indicatevariations thatwouldbeconsideredwithin the rangeofnatural variability, so inotherwords,notnecessarilytheresultofclimatechange. Theorangecolours indicateprojectiontimeserieswherethechangeswouldbeconsideredoutsideoftherangeofnaturalvariabilityandsolikelyaresponsetoclimatechange.It is importanttonotethattheseareGlobalClimateModel(GCM)projectionsandsolikelydonotcapturelocalscalefeaturessuchastopographyandland-oceanboundarydynamics.Theyalsomaynot capture small-scale features such as severe thunderstorms that can have important societalimpacts. Finally, these projections are averages over relatively large spatial area which differbetweenGCMsanditispossiblethatdifferentmessageswouldbeobtainedatsmallerspatialscalesandifvariousformsofdownscalingwereperformed.

Figure14:CMIP5projectedchanges inannualmeandailymaximumtemperatureunder theRCP8.5 concentrationpathway forDar es Salaam.Theblack line shows themulti-modelmean valueacrossallmodels in the referenceperiod1986-2005.Thecoloured lines showthe20-yearmovingaverageof results fromeachmodel and the shading around each line shows the 95% confidencerange around thosemodel results.Where the line and associated shading changes from blue tored/orange,thisindicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.

25

Figure15:CMIP5projectedchanges inannualmeandailyminimumtemperatureunder theRCP8.5 concentrationpathway forDar es Salaam.Theblack line shows themulti-modelmean valueacrossallmodels in the referenceperiod1986-2005.Thecoloured lines showthe20-yearmovingaverageof results fromeachmodel and the shading around each line shows the 95% confidencerange around thosemodel results.Where the line and associated shading changes from blue tored/orange,thisindicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.

Figure 16: CMIP5projected changes in annual hot days (dailymaximum temperature > 33.4 °C)undertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodels in the referenceperiod1986-2005.Thecoloured lines showthe20-yearmovingaverageofresultsfromeachmodelandtheshadingaroundeach lineshowsthe95%confidencerangearoundthosemodelresults.Wherethelineandassociatedshadingchangesfromblue to red/orange, this indicates when 20-year moving averages move outside of the 95%confidencerangeofthereferenceperiod.

26

Figure 17: CMIP5 projected changes in annual hot nights (dailyminimum temperature > 24 °C)undertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodels in the referenceperiod1986-2005.Thecoloured lines showthe20-yearmovingaverageofresultsfromeachmodelandtheshadingaroundeach lineshowsthe95%confidencerangearoundthosemodelresults.Wherethelineandassociatedshadingchangesfromblue to red/orange, this indicates when 20-year moving averages move outside of the 95%confidencerangeofthereferenceperiod.

Figure 18: CMIP5 projected changes in annual total rainfall under the RCP 8.5 concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodelsinthereferenceperiod1986-2005.Thecolouredlinesshowthe20-yearmovingaverageofresultsfromeachmodelandtheshadingaroundeachlineshowsthe95%confidencerangearoundthosemodelresults.Wherethelineandassociatedshadingchangesfrombluetored/orange,thisindicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.

27

Figure 19: CMIP5 projected changes in rainfall daily intensity under the RCP 8.5 concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodelsinthereferenceperiod1986-2005.Thecolouredlinesshowthe20-yearmovingaverageofresultsfromeachmodelandtheshadingaroundeachlineshowsthe95%confidencerangearoundthosemodelresults.Wherethelineandassociatedshadingchangesfrombluetored/orange,thisindicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.

Figure20:CMIP5projectedchangesinannualfrequencyofheavyrainfalldaysundertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodelsinthereferenceperiod1986-2005.Thecolouredlinesshowthe20-yearmovingaverageof results from each model and the shading around each line shows the 95% confidence rangearound those model results. Where the line and associated shading changes from blue tored/orange,thisindicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.

28

Figure21:CMIP5projectedchangesinannualfrequencyofextremerainfalldaysundertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodelsinthereferenceperiod1986-2005.Thecolouredlinesshowthe20-yearmovingaverageof results from each model and the shading around each line shows the 95% confidence rangearound those model results. Where the line and associated shading changes from blue tored/orange,thisindicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.StatisticallyDownscaledGlobalClimateModelsThe plots below (Figures 22 to 29) are called plume plots and they are used to represent thedifferent long-term projections across themultiple statistically downscaled climatemodels in theCMIP5modelarchiveused to informthe IPCCAR5 report. Theplots showprojectedvariations indifferentvariablesforthegridcelloverDaresSalaamproducedbyanensembleof11models.Thebluecoloursindicatevariationsthatwouldbeconsideredwithintherangeofnaturalvariability,soinotherwords, not necessarily the result of climate change. The orange colours indicate projectiontimeserieswherethechangeswouldbeconsideredoutsideoftherangeofnaturalvariabilityandsoarelikelyaresponsetoclimatechange.ItisimportanttonotethatthesearedownscaledGCMprojections,whichhaveaspatialresolutionofroughly50km.TheyprovidehigherresolutionoutputthantherawGCManddepictthefirstorderresponse to anthropogenic response. However they are unlikely to accurately capture local-scalefeaturessuchastopographyandland-oceanboundarydynamics.Theyalsomaynotcapturesmall-scalefeaturessuchasseverethunderstormsthatcanhaveimportantsocietalimpacts.

29

Figure22:StatisticallydownscaledprojectedchangesinannualmeandailymaximumtemperatureundertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodels in the referenceperiod1986-2005.Thecoloured lines showthe20-yearmovingaverageofresultsfromeachmodelandtheshadingaroundeach lineshowsthe95%confidencerangearoundthosemodelresults.Wherethelineandassociatedshadingchangesfromblue to red/orange, this indicates when 20-year moving averages move outside of the 95%confidencerangeofthereferenceperiod.

Figure23:StatisticallydownscaledprojectedchangesinannualmeandailyminimumtemperatureundertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodels in the referenceperiod1986-2005.Thecoloured lines showthe20-yearmovingaverageofresultsfromeachmodelandtheshadingaroundeach lineshowsthe95%confidencerangearoundthosemodelresults.Wherethelineandassociatedshadingchangesfromblue to red/orange, this indicates when 20-year moving averages move outside of the 95%confidencerangeofthereferenceperiod.

30

Figure24: Statisticallydownscaledprojected changes in the frequencyofdayswithTmax>90thpercentileofthehistoricalperiod(1986-2005)undertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodelsinthereferenceperiod1986-2005.Thecolouredlinesshowthe20-yearmovingaverageofresultsfromeachmodelandtheshadingaroundeachlineshowsthe95%confidencerangearoundthosemodelresults.Wherethelineandassociated shadingchanges fromblue to red/orange, this indicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.

Figure 25: Statistically downscaledprojected changes in the frequencyof dayswith Tmin > 90thpercentileofthehistoricalperiod(1986-2005)undertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodelsinthereferenceperiod1986-2005.Thecolouredlinesshowthe20-yearmovingaverageofresultsfromeachmodelandtheshadingaroundeachlineshowsthe95%confidencerangearoundthosemodelresults.Wherethelineandassociated shadingchanges fromblue to red/orange, this indicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.

31

Figure 26: Statistically downscaled projected changes in annual total rainfall under the RCP 8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodelsinthereferenceperiod1986-2005.Thecolouredlinesshowthe20-yearmovingaverageof results from each model and the shading around each line shows the 95% confidence rangearound those model results. Where the line and associated shading changes from blue tored/orange,thisindicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.

Figure 27: Statistically downscaled projected changes in the daily intensity of rainfall under theRCP 8.5 concentration pathway for Dar es Salaam. The black line shows the multi-model meanvalue across all models in the reference period 1986-2005. The coloured lines show the 20-yearmoving average of results from each model and the shading around each line shows the 95%confidencerangearoundthosemodelresults.Wherethelineandassociatedshadingchangesfromblue to red/orange, this indicates when 20-year moving averages move outside of the 95%confidencerangeofthereferenceperiod.

32

Figure28: Statisticallydownscaledprojected changes in the frequencyofheavy raindaysundertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalue across all models in the reference period 1986-2005. The coloured lines show the 20-yearmoving average of results from each model and the shading around each line shows the 95%confidencerangearoundthosemodelresults.Wherethelineandassociatedshadingchangesfromblue to red/orange, this indicates when 20-year moving averages move outside of the 95%confidencerangeofthereferenceperiod.

Figure29:Statisticallydownscaledprojectedchangesinthefrequencyofdayswithrainfall>90thpercentileofthehistoricalperiod(1986-2005)undertheRCP8.5concentrationpathwayforDaresSalaam.Theblacklineshowsthemulti-modelmeanvalueacrossallmodelsinthereferenceperiod1986-2005.Thecolouredlinesshowthe20-yearmovingaverageofresultsfromeachmodelandtheshadingaroundeachlineshowsthe95%confidencerangearoundthosemodelresults.Wherethelineandassociated shadingchanges fromblue to red/orange, this indicateswhen20-yearmovingaveragesmoveoutsideofthe95%confidencerangeofthereferenceperiod.

33

Acknowledgements WeacknowledgetheWorldClimateResearchProgramme'sWorkingGrouponCoupledModelling,whichisresponsibleforCMIP,andwethanktheclimatemodellinggroups(listedinTable4ofthisreport) for producing andmaking available theirmodel output. For CMIP theU.S. Department ofEnergy'sProgram forClimateModelDiagnosisand Intercomparisonprovidescoordinating supportandleddevelopmentofsoftwareinfrastructureinpartnershipwiththeGlobalOrganizationforEarthSystemSciencePortals

34

The contents of this Working Paper reflect the views of the author only and not those of the UK Department forInternationalDevelopmentortheEconomicandSocialResearchCouncil.