Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Lack of vegetation exacerbates exposure to dangerous heat in dense settlements in a tropical African city

Lack of vegetation exacerbates exposure to dangerous heat in dense settlements in a tropical... BothclimatechangeandrapidurbanizationaccelerateexposuretoheatinthecityofKampala, Uganda.Fromanetworkoflow-costtemperatureandhumiditysensors,operationalin2018–2019, wederivethedailymean,minimumandmaximumHumidexinordertoquantifyandexplain intra-urbanheatstressvariation.Thistemperature-humidityindexisshowntobeheterogeneously distributedoverthecity,withadailymeanintra-urbanHumidexIndexdeviationof1.2 Con average.Thelargestdifferencebetweenthecoolestandthewarmeststationoccursbetween16:00 and17:00localtime.Averagedoverthewholeobservationperiod,thisdailymaximumdifference ◦ ◦ is6.4 Cbetweenthewarmestandcooleststations,andreaches14.5 Conthemostextremeday. Thisheatstressheterogeneityalsotranslatestotheoccurrenceofextremeheat,showninother partsoftheworldtoputlocalpopulationsatriskofgreatdiscomfortorhealthdanger.Onestation inadensesettlementreportsadailymaximumHumidexIndexof >40 Cin68%ofthe observationdays,alevelwhichwasneverreachedatthenearbycampusoftheMakerereUniversity, andonlyafewtimesatthecityoutskirts.Largeintra-urbanheatstressdifferencesareexplainedby satelliteearthobservationproducts.NormalizedDifferenceVegetationIndexhasthehighest(75%) powertopredicttheintra-urbanvariationsindailymeanheatstress,butstrongcollinearityis foundwithothervariableslikeimpervioussurfacefractionandpopulationdensity.Ourresults haveimplicationsforurbanplanningontheonehand,highlightingtheimportanceofurban greening,andriskmanagementontheotherhand,recommendingtheuseofa temperature-humidityindexandaccountingforlargeintra-urbanheatstressvariationsand heat-pronedistrictsinurbanheatactionplansfortropicalhumidcities. 1.Introduction Schwartz2007,OudinÅströmetal2011,Fischeretal 2012, Mora et al 2017), raising serious concerns for Heat is a killer hazard with a global reach. Its expos- human health in a projected warmer future climate ure has been associated with both increased mortal- (Kovats and Hajat 2008, Huang et al 2011, Guo et al ity and morbidity worldwide (Medina-Ramón and 2014, Mora et al 2017). Former research on health ©2022TheAuthor(s). PublishedbyIOPPublishingLtd Environ. Res. Lett.17(2022)024004 JVandeWalleetal impactofextremeheatconcentratesonmid-latitude, intra-urbanvariations.LCZarethusexpectedtoalso high-income countries of low to medium popula- reflect heat stress variations in the city (Kabano et al tion density (Campbell et al 2018, Green et al 2019, 2021,VandeWalleetal2021),similartothefindings Otto et al 2020), thereby chronically underreporting in Nairobi (Kenya), concluding that informal settle- regions that are projected to actually experience the mentsareparticularlypronetoheatstress(Scottetal mostextremeheatinthefuture(Imetal2017,Mora 2017). etal2017,Nagendraetal2018,HarringtonandOtto However, observational studies investigating this 2020, Saeed et al 2021). For example, Africa is par- heterogeneity have been depreciated, because of the ticularly vulnerable to heat stress (IPCC 2014, Singh characteristic meteorological data scarcity in the et al 2019). A rapid increase in the intensities and region (Roth 2007). Six weather stations were set up frequenciesofheatwavesduringthepastdecadeshas inKampalaonlyrecently,thankstotheTrans-African beendemonstrated(Ceccherinietal2017,Amouetal Hydro-Meteorological Observatory (TAHMO, van 2021), while simulations project this trend to con- de Giesen et al 2014) project, collecting meteor- tinueuninterruptedintothefuture(Harringtonetal ological data from the synoptic station at Maker- 2016,Russoetal2016,Dosioetal2018).Forinstance, ere university and five instrument shelters placed in underahigheremissionscenario(SSP5-8.5),Africa’s open school gardens, in accordance with the offi- exposure to extreme heat is projected to be 7–269 cial World Meteorological Organization standards times larger than it has been historically (Liu et al (WMO1986).Despitethisgreatobservationaleffort, 2017,Asefi-Najafabadyetal2018). nostationsareplacedinmoredenselybuiltenviron- Extremeheatis furtheramplified incities, which mentswheremostofKampala’spopulationlives. areshowntobewarmerthantheirnaturalsurround- To better represent the variations of heat stress ings, because of reduced vegetated areas, increased throughout the city of Kampala, including densely release of anthropogenic heat, changes in surface populated areas, this study put in place an obser- albedo and trapped radiation within street canyons vational network of 45 low-cost iButton sensors. (Oke 1982). This urban heat island effect has also These sensors recorded near-surface air temperat- been demonstrated for Sub-Saharan African cities ure and relative humidity for three 42 d periods (Nakamura 1966, Jonsson et al 2004, Roth 2007, between August 2018 and April 2019 (Van de Walle Brousse et al 2020), experiencing rapid population et al 2021). From these measurements, the Humi- growth(McGranahanandSatterthwaite2014,United dex Index is computed, providing a good estimate Nations 2019) and extensive urbanization (Liu et al for feel-like temperature (Masterton and Richard- 2017, United Nations 2018, Marcotullio et al 2021). son 1979). High relative humidity decreases a per- As an example, Kampala, the capital city of Uganda, son’s evaporation ability and thereby the effective- is experiencing an uncontrolled urbanization, hav- ness of the body’s natural cooling system (Malchaire −1 ing the fourth highest growth rate (>4%yr ) of all et al 2000, Hass et al 2016). Particularly in hot and African cities (Richmond et al 2018, Kampala Cap- humid cities like Kampala, high values of the Humi- ital City Authority and Uganda Bureau of Statist- dex Index might cause dangerous health conditions. ics 2019). Like many fast-growing cities, Kampala is We therefore focus on extreme heat recorded at the expanding horizontally (Brousse et al 2019, Li et al different stations, and explain the observed patterns 2021),demonstratingspatialpatternsofurbansprawl basedonrelevantsatellite-derivedearthobservations. (Vermeiren et al 2016, Hemerijckx et al 2020) and For example, vegetation is known to generally play the formation of informal settlements or slums (Van a twofold role, decreasing temperature but enhan- Leeuwenetal2017,Lwasaetal2018,Richmondetal cing humidity by transpiration (Hass et al 2016). 2018). Results are discussed from two different perspect- Within the city of Kampala, both morpholo- ives:insightsinspatialheterogeneityofheatinKam- gicalandsocio-economicalcharacteristicslargelydif- pala and occurrences of heat above great discomfort fer, distinguishing wealthy districts characterized by thresholdsamongdifferenturbanenvironments. asphalted roads, modern houses and large gardens, frominformalsettlementscomposedofdenselybuilt 2.Methods shacksmadeofcorrugatedmetalsheetsthatareonly accessible via small alleys (Vermeiren et al 2012, 2.1.iButtonobservations Hemerijckxetal2020).Recently,Brousseetal(2019) The iButton sensor, a product of Maxim Integrated, classifiedtheseintra-urbanvariationsintoLocalCli- is a low-cost sensor containing a temperature and mateZones(LCZ,StewartandOke2012).Thisclassi- humidityloggingsystem(Hubbartetal2005).Witha ficationincludes7vegetatedand10builtclasses,each loggingfrequencyanddataaccuracyprogrammedat class exemplifying uniform surface cover, structure, 15 min and 11 bit respectively, each sensor can store material and human activity that span hundreds of 42 consecutive days of data. Afterwards, a manual meterstoseveralkilometersinhorizontalscale(Stew- download is required. To protect the sensors from art 2011). Importantly, LCZs are designed to reflect radiationandsplashwater,theyareshieldedbyafol- the thermal environments as a consequence of their dedthinlightreflectivefilm(figureS1availableonline 2 Environ. Res. Lett.17(2022)024004 JVandeWalleetal Figure1.LocationsofthedifferentsensorsthroughoutthecityofKampala,withtopviewoftheneighbourhoodsaroundthe stationsNamungoona(Ng),Makerere(Mk),Nakasero(Ns),Industrialarea(Ia),Nkeere(Nk)andBuziga(Bz).Asetofthree sensorsisinstalledperlocation(attheexactcentreofeachtopviewimage),reducingtheuncertaintyduetodifferentinstallation conditions.ExactlocationsofthesensorsarelistedintableS1.Eachtopviewimageis500 ×500m ,retrievedfromGoogleEarth imagery.Thelandsurfacetemperatureestimationforthecloud-freedayof27February2021isderivedfromLandsat8and MODISremotesensingproductsviaParastatidisetal(2017),andshowslargeintra-urbanvariationsofsurfacetemperatures. at stacks.iop.org/ERL/17/024004/mmedia), designed values. These periods do not cover a full year, yet we and produced by the Maryland Institute College monitored humidity and temperature for both dry of Art and Johns Hopkins University. The sensor- andwetseasons.Theintra-seasonalvariationsintem- containingshieldsarezip-tiedpreferablytoawooden peratureandhumidityaregenerallyratherlimitedin material,atabout2mheightandatashadedlocation the region (figure S3). On a longer timescale, warm (tableS1).Inthisstudywereusedthematerialofthe spellscanoccur,withthestartof2019asanexample. observational campaign by Scott et al (2017) held in The second of our observational periods covers that Nairobi,Kenya. period (figure S3). Due to a slightly different down- The resulting network of 45 sensors installed loading moment, measurement periods may slightly throughout the city of Kampala (figure 1) aimed at differ for locations located far from each other. In properlyrepresentingthecity’ssurfaceheterogeneity. addition, some data is missing for the second and Besides a good spatial coverage of both openly and third periods due to technical issues, especially low denselybuiltenvironments,thechoiceofsensorsloc- battery. At the end of the third period, 32 out of 45 ationalsoconsideredthesecurityofthesensorsfrom sensorsremainedactive. vandalism. Preferred locations thus included schools In addition, one sensor set was put indoor at the or houses of local acquaintances. At each location, a informal settlement of Acholi Quarters. The build- setofthreesensorswereinstalledclosetogether,redu- ing is a concrete block, no windows, an iron door cingtheuncertaintyofthetriple-sensor-meandueto andironroofsheets.Giventhefactthatitisonlyone different installation conditions such as attachment site, results are qualitatively described in the discus- material, shade fraction or ground cover. For each sionsection. location, the 15 min resolution triple-sensor-mean 2.2.Measurementquality information is reduced to the minimum and max- imum values per day. In addition, the daily mean is Themanufacturer’sevaluationoftheiButtonsreports a thermochron and hygrochron accuracy of 0.5 C computedastheaverageover24h.Threedownload- ing rounds provided data for 3 periods of 42 d each and 5% respectively. This is confirmed by calculat- ing the mean deviation per sensor triple, explained betweenAugust2018andApril2019(figureS2).For furtheranalyses,dailyminimumanddailymaximum and summarized in table S2. In addition, the triple-sensor-meanisevaluatedagainsttheMakerere valuesarecomputedastheaveragesoverall3 ×42d, stillreferredtoasdailyminimumanddailymaximum automatic weather station data, part of the TAHMO 3 Environ. Res. Lett.17(2022)024004 JVandeWalleetal network (van de Giesen et al 2014). With an overall vegetationcancounteracttheurbanheatislandeffect, ◦ ◦ ◦ temperature bias of 0.10 C, 0.11 C and 0.53 C for example by evaporative cooling (Oke 1982). The in the three periods, and relative humidity bias of normalized digital vegetation index (NDVI) ranging −2.50%, −2.66% and −1.43%, the iButton sensors from0to1,isusedasaproxyforthefractionofsur- tend to slightly overestimate the air temperature face covered by vegetation. This product is derived and underestimate humidity. This however varies from the Moderate Resolution Imaging Spectrora- throughout the day as observed nighttime temperat- diometer (MODIS) onboard the Terra satellite with ures by the iButton sensors are higher than the ones a horizontal resolution of 250 m. This satellite over- measured by the automatic weather station, while passesKampalaat10:30AMandPMlocaltime.Our observeddaytimetemperaturesarelower(figureS4). analysis uses the median value of two years, 2018– This results in an underestimation of the diurnal 2019, overlapping the observation period. No tem- temperature range. Nighttime relative humidity is poral variation is taken into account, assuming little underestimated by ∼4%, but daytime observations seasonal variation in this tropical area and keeping compare well to the Makerere automatic weather the focus to the spatial heterogeneity. Fourth, due station. to the high heat storage capacity of building mater- ials, densely built areas can largely influence local 2.3.Humidexindex temperatures (Oke 1973). As a proxy for this build- To better estimate the human-experienced heat, the ing density, impervious surface area (ISA) fraction Humidex Index (hereafter referred to as ‘Humi- is retrieved from the Global Man-made Impervious dex’,H)iscomputedevery15minfromobservations Surface(GMIS)dataset(DeColstounetal2017).For of both temperature (T in C) and relative humid- the target year 2010, the GMIS product analysed all ity (RH in %), following Masterton and Richardson cloud-free images from Landsat 5 and 7, inheriting (1979): thehighhorizontalresolutionof30m.Astrongcor- ( ) relation is expected, yet the abundance of bare soil 7.5 T 5 RH 237.7+T H(T,RH) =T + 6.112 10 −10 . oftenchallengessatelliteinstrumentstoproperlyrep- 9 100 resentISA(VandeWalleetal2021).Fifth,anthropo- (1) genicheat,mainlyfromdomesticandtransportation fueluse,isproducedinhighlypopulatedareas(Taha The resulting quantity increases non-linearly with 1997, Stewart and Kennedy 2015). The population both air temperature and relative humidity, and can densityofthegreaterKampalaregion,formallyavail- be understood as feel-like temperature in degrees ableperdistrictfortheyear2014(UgandaBureauof Celsius. Humidex values above 40 C lead to ‘great Statistics 2014, Hemerijckx et al 2020), is translated discomfort’, values exceeding 45 C are ‘dangerous’ to a 30 m resolution grid. Sixth, the MODIS instru- (MastertonandRichardson1979).Humidexinform- mentalsoprovidesdirectionalhemispherical(black- ationisreducedtotheminimumandmaximumval- sky) near-infrared albedo at 0.7–5.0 µm wavelength uesperday,aswellasthedailymeanwhichistheaver- at 500 m horizontal resolution, possibly distinguish- ageover24h. ingdifferentroofingtypeswithinthecityofKampala (Brest1987). 2.4.Explanatoryvariables IfthesevariablescanexplaintheobservedHumi- WeaimtoexplainspatialHumidexpatternsbycom- dexvariations,asimplestatisticalmodelcouldextra- paring them against potential explanatory factors, polate operational weather station data to the entire including distance to the lake, vegetation presence, city, providing heat stress information about hardly built-upfraction,surfaceelevation,populationdens- accessible locations such as informal settlements. ityandsurfacealbedo(figureS5).Thechoiceofeach Assuming linear behaviour, a multiple linear regres- ofthesesixfactorsisarguedbelow.First,locatednext sion technique is applied, expressing the Humidex to Lake Victoria, Kampala is affected by a daytime (H)intermsoftheexplanatoryvariablesx : lake breeze developing on a daily basis (Thiery et al 2015, 2016, 2017, Brousse et al 2020, Van de Walle et al 2020, Woodhams et al 2021). The distance of H = β + βx , (2) 0 i i thesensorstothelakeisthereforeusedtoaccountfor i=1 different onsets of cooling. Second, hills can optim- allybenefitfromcoolingwinds,lowerregionscannot. where the Ordinary Least Squares method estim- In addition, these low areas are typically wetlands, ates the best fitting coefficients β based on the providing opportunities for urban farming, but also observations.Initially,allN explanatoryvariablesare humidifying the air (Kabumbuli and Kiwazi 2009). included,butat-testdecidesontheeliminationofthe Therefore, a digital elevation model at 30 m hori- least significant variable. This backward elimination zontal resolution is retrieved from the Shuttle Radar continues iteratively until all explanatory variables Topography Mission (SRTM, Farr et al 2007) as a are significant at 95% confidence level. Ultimately, potentialexplanatoryvariable.Third,thepresenceof the remaining multiple linear regression model no 4 Environ. Res. Lett.17(2022)024004 JVandeWalleetal longer suffers from multicollinearity, ensuring its TheobservedHumidexheterogeneityisaresultof coefficientstobeoptimallystable(HalinskiandFeldt intra-urban temperature and relative humidity vari- 1970). ations.Thelatterareconsideredattimesofdailymin- imumandmaximumHumidex(figures2(a),(d)and (c), (f) respectively). Around sunrise, when Humi- 2.5.Localclimatezones dexreachesitsminimum,theairtemperaturealmost In addition to this quantitative approach, locations entirely determines the Humidex variations between are classified based on the building structure, land the sensor sites, with the urban air being clearly cover and human activity according to the LCZ clas- hotter and drier than at the outskirts (figures 2(a), sification scheme (Stewart and Oke 2012, Brousse (d) and (g)). Specific humidity between those sites etal2019,seetableS1).TheoutskirtsstationsBuloba is similar (not shown). Around noon, when Humi- (Bl), Kawanda (Kw), Bukerere (Bk) and Buziga (Bz) dex reaches its maximum, the situation is differ- are located in open low-rise environments (LCZ 6), ent. Then, the highest air temperatures are observed defined by small (3–10 m) buildings with abund- at Nateete (Nt), Nkeere (Nk) and Acholi Quarters ant plant cover (Stewart 2011). The campus of the (Aq,figure2(c)),allclassifiedaslightweightlow-rise Makerere University (Mk) is classified as open mid- LCZ 7. These locations also have the lowest relative rise class (LCZ 5), characterized by open arrange- humidity(figure2(f)).WhilethetemperatureatNaj- ment of 3–9 story buildings and abundant plant janankumbi(Nj,compactlow-riseLCZ3)isverysim- cover. The Industrial Area (Ia) station is located ilar to the temperatures at Nateete or Nkeere, it has in a large low-rise (LCZ 8) class, characterized by a clearly higher Humidex (figure 2(i)). This can be extensivepavedsurfacesbetweenlarge,lowbuildings, explained by Najjanankumbi’s high relative humid- oftenwithanindustrialorcommercialfunction.The ity compared to Natalee and Nkeere. We therefore Nakasero(Ns)stationislocatedincompactmid-rise needbothtemperatureandrelativehumiditytoprop- (LCZ 2) class, defined by buildings of 10–25 m sep- erly explain spatial variations in heat. Importantly, arated by narrow streets and inner courtyards, and neither Najjanankumbi’s air temperature or relative with few or no trees. In addition, we classify the sta- humidity is exceptional compared to other stations tions in Namungooma (Ng), Bwaise (Bw) and Naj- such as Nkeere or Makerere: the temperature distri- janankumbi (Nj) as compact low-rise (LCZ 3) and bution is similar to the one at Nkeere (figure S6(c)), Nateete(Nt),Nkeere(Nk)andAcholiQuarters(Aq) and relativehumidity valuesexceeding 70% occur as aslightweightlow-rise(LCZ7),oftencalledinformal frequent at Makerere (figure S6(f)). Instead, it is the settlements or slums. Both classes consist of small combination of both compound drivers that creates buildingstightlypackedalongnarrowstreetswithno high maximum Humidex values in Najjanankumbi or little vegetation. Typical for the latter class are the (figureS6(i),Zscheischleretal2018,2020). lightweight building materials (thatch, wood or cor- Relating the observed Humidex heterogeneity to rugatedmetal)andoftenformlessarrangementofthe the LCZ classification, the open low-rise (LCZ 6) buildings(Stewart2011). environments, together with the open mid-rise Makerere University campus (LCZ 5), report gen- erally lowest temperature, highest relative humidity 3.Results and lowest Humidex values (figure 2). These cool Measurements show clear differences in Humidex environments contrast with the warm compact and values at the different sensor locations (figures 2(g)– lightweight low-rise classes (LCZ 3 and LCZ 7), as (i)). For example, the average Humidex varies well as the compact mid-rise central business centre ◦ ◦ between 30.6 C and 32.1 C for the urban stations, (LCZ 2). Kampala’s industrial area (Ia, LCZ 8) is exceptfortheMakerere(Mk)station,locatednearthe rather warm at night, but shows relatively mild tem- city centre, with a substantially lower average Humi- peraturesduringdaytime.Especiallypoorlyvegetated dex (29.0 C). Also the city’s outskirts, represented andcompactlybuiltneighbourhoodsinKampalaare by Buloba (Bl), Bukerere (Bk) and Kawanda (Kw) thus more prone to heat stress than the outskirts or ◦ ◦ stations, are cooler (29.3 C–29.9 C, figure 2(h)). Makererestation. At night (figure 2(g)), both the city outskirts and Intra-urban differences are especially relevant Makerere are cool, in contrast with high Humidex whenconsideringextremeheat(figure3).Concretely, values observed nearby central stations. Particularly Makerere (Mk) has no single day with the Humidex theindustrialarea(Ia),Nakasero(Ns),Namungoona exceeding the ‘great discomfort’ threshold of 40 C, (Ng)andNkeere(Nk)experiencewarmnights,with itoccursin2%–16%oftheobserveddaysintheout- differences up to 2.3 C compared to Makerere. skirtsstationsBukerere(Bk),Kawanda(Kw),Buloba The intra-urban heterogeneity is most pronounced (Bl) and Buziga (Bz), and 14%–67% in the stations when comparing daily maximum Humidex values locatedindenselybuiltareas,inparticularinNkeere (figure 2(i)), ranging from 34.1 C at Makerere to (Nk),AcholiQuarters(Aq)andNajjanankumbi(Nj) 41.6 CatNajjanankumbi. (in50%–67%oftheobserveddays).Lookingatdays 5 Environ. Res. Lett.17(2022)024004 JVandeWalleetal Figure2.Observedtemperatureandrelativehumidityatalllocations,aswellastheHumidexderivedfollowingequation(1).The leftandrightcolumnsshowobservationsatdailyminimumandmaximumHumidexrespectively,whilethemiddlecolumngives theHumidexaveragedoverthethreeobservationperiods(3 ×42d).ThegreycontourlineindicatestheKampalaurbanextent, definedbyISAfractionabove0.1(figureS5).DifferentmarkersymbolscorrespondtothedifferentLCZclassesobtainedfrom Brousseetal(2019)andVandeWalleetal(2021). with maximum Humidex exceeding 45 C, the ‘dan- correlations are found between Humidex and the gerous’threshold,Bwaise(Bw),Najjanankumbi(Nj) proximity to the lake or elevation. Importantly, the and to a lesser extent Nkeere (Nk) stand out, with explanatoryvariablesarenotindependentfromeach occurrences of 17%, 12% and 4% of the observed other, with high ISA fractions prohibiting abundant days,respectively. vegetation(correlationof −0.93)whilealsoimplying Not only for daily maximum Humidex, also lower near-infrared albedo due to the strongly mod- for daily minimum or average Humidex, Nkeere ified land cover (correlation of −0.82, figure S10). (Nk), Acholi Quarters (Aq) and Namungoona (Ng) Also population density is not independent from show highest exceedance frequencies of all Humi- ISA fraction, NDVI or near-infrared albedo, with dexthresholds(figureS7).Whenalsoaccountingfor correlations of 0.71, −0.78 and −0.72, respectively. the duration of exceedance by considering the hours The elevation shows a moderate correlation of 0.51 above a certain threshold (figure S8, top), or for the with NDVI, probably related to Kampala’s vegetated heat intensity by computing the mean heat-degree- hills. hours(figureS8,bottom),thedenselybuiltenviron- Due to this collinearity between the explanat- mentsoftenexperienceextremeheat(1%–10%ofthe oryvariables,thestepwisebackwardeliminationpro- observedtime),especiallyinNajjanankumbi(Nj). cedure only retains NDVI as explanatory variable The intra-urban heterogeneous Humidex values in the linear regression model for minimum, mean correlates with six proposed explanatory variables and maximum Humidex (figure 4). With R =0.79, (figureS9).Strongestcorrelationsarefoundbetween the NDVI has a high explanatory power for min- Humidex and NDVI as well as ISA fraction. Moder- imum (early morning) Humidex, meaning that 79% ate negative correlations of −0.5 to −0.65 are found of the the intra-urban Humidex variability can be between Humidex and near-infrared albedo, while explained.Thisexplanatorypowerissimilarforaver- Humidex and population density are positively cor- age(75%),butlowerformaximum(midday)Humi- related, with values between 0.5 and 0.76. Smaller dex(52%). 6 Environ. Res. Lett.17(2022)024004 JVandeWalleetal Figure3.Exposuretimetoheatstressextremesforeachlocation.Timeisgiveninpercentageofobserveddaysoutof3 ×42dat whichthedailymaximumHumidexexceedsacertainthreshold,definedbythevaluesontheabscissa.Typicalthresholdsare ◦ ◦ ◦ indicated,relatedto‘somediscomfort’ifH >30 C,to‘greatdiscomfort’ifH >40 Candto‘dangerous’ifH >45 C. max max max Figure4.(b)DependencyoftheHumidexaveragedovertheperiodof3 ×42donvegetationfraction(NDVI)foreachlocation. Samefordependenciesofdailyminimum(a)andmaximum(c)Humidex.ColourscorrespondtotheLCZclassesassignedbased ontheirGoogleEarthtop-viewimages(seefigure1).Linearregressionresultsafterastepwisebackwardeliminationprocedureof otherexplanatoryvariablesisshownbythefullbluelineandthecorrespondingequation,whiledashedlinesdefinethe95% 2 2 confidencebands.TheR valuesprovidetheexplanatorypoweroftheregressionmodels,withR adjustingforthenumberof adj predictors. Figure5.Extrapolationofdailyminimum(a),mean(b)andmaximum(c)Humidexbasedontheregressionmodelsfrom figure4.Inpractice,theHumidexisderivedfromtheexplanatoryvariableNDVIavailableat250mresolution(figureS5), supplementedbyandmultipliedwithtwocoefficients β and β (equation(2))resultingfromthelinearregressionprocedure. 0 1 The explanatory power of NDVI allows us to 4.Discussion applytheregressionmodelbyextrapolatingthepoint observations of minimum, mean and maximum Withaverageintra-urbandifferencesof1.2 C,andan Humidex to the greater Kampala region (figure 5). afternoondifferenceof6.4 Conaverage,theHumi- Thismapprovidesinformationonthespatialhetero- dex Index is heterogeneously distributed over the geneityofheatstressinKampala. city of Kampala. These large intra-urban differences 7 Environ. Res. Lett.17(2022)024004 JVandeWalleetal are also reflected in the number of days exposed to open low-rise (LCZ 6) and open mid-rise (LCZ 5) extremeheat.Atsomelocations,dailymaximumheat classes. Overall, sparsely built and highly vegetated stress exceeds the great discomfort level, defined by areasthusexperiencesubstantiallylowerheat,withan ◦ ◦ Humidex >40 C, for more than 50% of the 3 × observedmeandifferenceof6.4 Cintheafternoon. 42 observation days. In comparison, for the same This result implies that greening the city could mit- period, this level was never reached at the Maker- igate urban heat (Bowler et al 2010, Demuzere et al ere station and only a few times in the city outskirts. 2014, Gunawardena et al 2017). Yet, a more detailed Moreover, we identified the regions in Kampala that investigationisneededtotheoveralleffectsofdiffer- are most prone to heat, pointed to non-vegetated, enttypesofvegetationonhumanwell-being,includ- densely built environments using linear regression ing the effects on local climate, air quality and aes- and extrapolated the result to the greater Kampala thetics (Salmond et al 2016). In fact, the Kampala region.Theresultingmapcouldcomplementremote City Council Authority (KCCA) announced plans to sensingproductsforlandsurfacetemperature,adding plant0.5milliontreesinKampalaaspartofitsclimate information about 2 m temperature and relative action strategy, which is developed with stakehold- humidity combined in the Humidex Index. Such ers. Concretely, the strategy has embarked on tak- map can guide anticipatory action plans that help ing stock of the trees in the city coupled with map- reduce the impact of heatwaves, by providing con- ping of natural assets in the city to form the basis crete information about heat-prone areas. With spe- for implementation of the climate strategy. This cli- cialattentiontothoseheat-proneareas,aheataction mateactionstrategyischallenging,especiallybecause plan commits to public awareness about heat risks most available land is in Kampala’s residential areas (Singh et al 2019), which has been demonstrated to where urban tree canopies are already evident, while be successful in reducing heat-related mortality in we showed that the need for cooling is most urgent Ahmedabad, India (Knowlton et al 2014, Hess et al in densely built environments and informal settle- 2018, Nastar 2020). In addition, a heat action plan ments. Yet, Lwasa et al (2014) claimed the potential also accounts for the vulnerability of urban dwell- forexpansionoftreecanopycoverinthecity’sdensely ers.Ingeneral,denselybuiltandinformalsettlements built and most vulnerable areas as well. For this, housealargepartofthepopulationbelongingtothe initiatives from local actors and inhabitants should lower socioeconomic status with income and liveli- be strongly supported, which can only be achieved hoodinsecurity,makingthemparticularlyvulnerable by properly informing the inhabitants (Hintz et al (Vermeiren et al 2012, Lwasa et al 2018, Hemerijckx 2018), followed by incentive-based mechanisms that et al 2020, Twinomuhangi et al 2021). An import- allow the planting, nurturing and taking care of the ant factor increasing their vulnerability is the hous- trees in dense settlements by the developers. As an inginfrastructure,notofferinganyprotectiontoheat. example,somelocalpoliticiansandinfluencerschal- Asatestcase,wecollectedheatobservationsinsidea lengedUgandansvia(social)mediatocommemorate building in the informal settlement of Acholi Quar- everymarriage,death,birthandgraduationbyplant- ters during the three periods of the observational ingapersonaltree. campaign.Insteadofofferingprotectionagainstheat, Fiveimportantchallengeshavetobeaddressedby the house seems to act like a heat trap for even- future research. First, besides Humidex, many other ingandnighttimeheat,timeswhenpeopleareliving indices can describe heat as a combination of mul- inside.Onlyheatbeforeandatnoonisslightlylower tiplefactorsinfluencingheatstress.Toexplainhuman thenoutsideobservations(figureS11).Arecentstudy discomfort, many factors play a role, including air in Ghana investigated indoor temperatures and con- temperatureandhumidity(EpsteinandMoran2006, cludedlargeeffectsofbuildingmaterials(Wilbyetal Barnett et al 2010, Fischer et al 2012, Lange et al 2021). 2020),butalsowind,radiation,physiologicalfactors, A second implication concerns urban adaptation physical activity and clothing (Quayle and Doehring planning, and follows from our finding that Kam- 1981, Roth 2007, Potchter et al 2018). This study pala’s intra-urban varying heat is strongly correl- includestemperatureandrelativehumidityonly,with ated with NDVI, explaining up to 77% of the intra- differences between similar heat measures shown to urban variability in daily mean Humidex. Despite be small (Barnett et al 2010). Second, to explain the their elimination in the regression analysis due to observed Humidex values, vegetation is an import- clearcorrelationswithNDVI,otherexplanatoryvari- ant factor with a twofold role. On the one hand, ables,particularlyISAfraction,mightalsobeimport- enhancedtranspirationbyvegetationimpliesairtem- ant. From a Local Climate Zone perspective (Stew- perature cooling, tending to reduce the Humidex. art and Oke 2012), the warmest stations were found Ontheotherhand,enhancedtranspirationincreases in compact LCZ classes, characterized by densely the humidity which contributes to a higher Humi- built environments with little or no vegetation. In dex (Hass et al 2016). By showing a strong negative particular, they include compact mid-rise (LCZ 2), correlation between Humidex and vegetation frac- compact low-rise (LCZ 3) and lightweight (LCZ 7) tion (NDVI), we conclude that the first role dom- classes. Cooler environments in Kampala include inates. Concretely, adding vegetation might increase 8 Environ. Res. Lett.17(2022)024004 JVandeWalleetal the relative humidity, the reduction in temperat- In some areas, heat is experienced as ‘great discom- ure is superior. Yet this study did not quantitatively fort’below/abovetheproposed40 Cthreshold.Sur- investigate the interplay between temperature and veys in Dar es Salaam (Tanzania) suggested that the humidity in detail. A study in Indiana, US, high- comfort range was well above the one defined in lighted the importance of different vegetation types temperate climates (Ndetto and Matzarakis 2017). and structures to unravel this interplay (Souch and Whileconcreteinformationaboutapplicablelevels/- Souch 1993). Individual tree species or open grass thresholds for Kampala is lacking, this study provi- areas have the largest cooling effect, street trees the sionally explored different thresholds for the Humi- ◦ ◦ smallest. Also in Kampala, different vegetation types dexIndex,rangingfrom30 Cto50 C.Fifth,though could be obtained or derived from detailed land- the explanatory power of the regression model is covermaps.Evenifsuchland-covermapswouldhelp high(75%),fourfactorscouldleadtoimprovements the interpretation of the results by further linking of the model. First, it would definitely benefit from heat to different vegetation types, our study already moreobservations,perhapssupplementedbyremote integrates the level of greenness and vegetation by sensing data. Second, this study explored only six using NDVI and ISA variables. Third, our observa- explanatory variables, but more could be added, for tional iButton network was only active for a short 3 exampleprovidinginformationaboutbuildingmor- × 42 d period, during a warm spell. Results might phology or materials. Though this information is therefore slightly overestimate the multi-year situ- implicitlyalreadyincludedinLCZclasses,highresol- ation. In addition, our network was not complete ution material maps will possibly appear in the near to cover all neighbourhoods. Moreover, paper-made future. Third, future research should investigate the shields might not be sufficient to perfectly protect robustness of the model when adding new observa- the sensor from direct sunlight. However, the role of tions. Fourth, we found strong collinearity between irradiation of the sensor shields is not investigated different explanatory factors, challenging the causal- since the time of the day at which the sensor is sun- ity between vegetation fraction and Humidex Index lit/shaded was not inventorized. Deviations between inourlinearregressionmodel. the sensor triple observations are small, suggesting that the effect of irradiation is also small. Despite 5.Conclusion thegoodperformanceofoursensors,westillrecom- mend using a pair or triplet of iButtons per site, Fromanetworkoflow-costtemperatureandhumid- regardless of the reduced number of observational ity sensors, we compute the Humidex Index, quanti- sites. Besides our iButton network, weather station fying the heat stress throughout the city. Daily min- observationsareincreasinglyconductedintheregion imum,mean,maximumaswellasextremeheatstress via the TAHMO project (van de Giesen et al 2014), are heterogeneously distributed over the city, with yetthemeteorologicalnetworkisstillsparse.Besides poorly vegetated and densely built-up environments TAHMO, the availability of crowd-sourced weather being the most heat-prone areas. Their inhabitants, data has grown worldwide in recent years, offering generally vulnerable people due to their socioeco- possibilities for high-resolution observational urban nomic status, are often exposed to great discom- heat studies (Muller et al 2015, Chapman et al 2017, fort or even dangerous heat. Future research should Meier et al 2017, Venter et al 2020, Fenner et al bridge the gap to indoor heat and its dependence on 2021). Unfortunately, no studies have collected and house structure and building materials. Two recom- analysed such data in tropical Africa. Fourth, there mendations follow from this study: to mitigate heat is no common definition of heatwaves (Hintz et al stress,urbangreeningshouldbeconsideredinurban 2018). This study investigated different definitions planning strategies, and urban heat action plans including total days/nights or hours with extreme should account for the large intra-urban heat stress heat and heat degree hours. Similar to Europe, the variations. US and Australia, a region-specific heatwave defini- tion should be derived from mortality and morbid- ity studies (Kovats and Hajat 2008, Guo et al 2014, Dataavailabilitystatement Mora et al 2017, Xu and Tong 2017, Li et al 2020). Yet such studies are currently completely lacking The data that support the findings of this study for Africa (Harrington and Otto 2020). In addition, are openly available at the following URL/DOI: given an appropriate metric, health impact research http://doi.org/10.5281/zenodo.5105570. should also question whether certain thresholds are Other products are publicly available online: the meaningful for different regions of interest based TAHMOautomaticweatherstationdataatMakerere on more metabolic information on people living in University (https://tahmo.org/climate-data/), SRTM these regions. In this study we provided the gen- Digital elevation model (https://gee.stac.cloud/6FKB eral (dis)comfort levels accompanying the Humidex uAFUoXyYMZCUkttkJ5VuS8cQd?t = bands), Index. Yet, experienced heat is expected to depend MODIS MYD13Q1 product for NDVI (https://mo on the region of the world (Potchter et al 2018). dis.gsfc.nasa.gov/data/dataprod/mod13.php), GMIS 9 Environ. Res. Lett.17(2022)024004 JVandeWalleetal built-up fraction (https://sedac.ciesin.columbia.edu/ Asefi-NajafabadyS,VandecarKL,SeimonA,LawrencePand LawrenceD2018Climatechange,populationandpoverty: data/set/ulandsat-gmis-v1), Kampala population vulnerabilityandexposuretoheatstressincountries density (https://unstats.un.org/unsd/demographic/ borderingthegreatlakesofAfricaClim.Change148561–73 sources/census/wphc/Uganda/UGA-2016-05-23.pdf) BarnettAG,TongSandClementsAC2010Whatmeasureof and MODIS MCD43A3 v6 albedo product (https:// temperatureisthebestpredictorofmortality?Environ.Res. 110604–11 modis.gsfc.nasa.gov/data/dataprod/mod43.php). In BowlerDE,Buyung-AliL,KnightTMandPullinAS2010Urban addition,thelandsurfacetemperatureisderivedfrom greeningtocooltownsandcities:asystematicreviewofthe Landsat 8 and MODIS, and can be retrieved online empiricalevidenceLandscapeUrbanPlan.97147–55 via:http://rslab.gr/downloads_LandsatLST.html BrestCL1987Seasonalalbedoofanurban/rurallandscapefrom satelliteobservationsJ.Appl.Meteorol.Climatol.261169–87 BrousseOetal2019UsinglocalclimatezonesinSub-Saharan Acknowledgments AfricatotackleurbanhealthissuesUrbanClim.27227–42 BrousseO,WoutersH,DemuzereM,ThieryW,VandeWalleJ andvanLipzigNP2020Thelocalclimateimpactofan ThisworkwasfinanciallysupportedbyInternalSpe- Africancityduringclear-skyconditions—implicationsof cialResearchFund(BOF-479C1ProjectC14/17/053) therecenturbanizationinKampala(Uganda)Int.J. and the REACT project. M Demuzere is suppor- Climatol.404586–608 ted by the ENLIGHT project, funded by the Ger- CampbellS,RemenyiTA,WhiteCJandJohnstonFH2018 Heatwaveandhealthimpactresearch:aglobalreviewHealth man Research Foundation (DFG) under Grant No. Place53210–8 437467569. The authors thank Daniel S Ddumba CeccheriniG,RussoS,AmeztoyI,FrancescoMarcheseAand (Makerere University) and Sadam Y Kiwanuka Carmona-MorenoC2017HeatwavesinAfrica1981–2015, (KCCA) for their support during the field work, and observationsandreanalysisNat.HazardsEarthSyst.Sci. 17115–25 local people for taking care of the sensors. OB was ChapmanL,BellCandBellS2017Canthecrowdsourcingdata financed by the Remote Sensing for Epidemiology paradigmtakeatmosphericsciencetoanewlevel?Acase in Sub-Saharan African CiTies (REACT http://react. studyoftheurbanheatislandofLondonquantifiedusing ulb.be/)project,fundedbytheSTEREO-IIIprogram NetatmoweatherstationsInt.J.Climatol.373597–605 DeColstounECB,HuangC,WangP,TiltonJC,PhillipsJ, of the Belgian Science Policy (BELSPO, SR/00/337) NiemczuraS,LingP-YandWolfeR2017Documentationfor andtheWellcomeTrustHEROICproject(GrantNo. theGlobalMan-madeImperviousSurface(GMIS)Dataset 216035/Z/19/Z). FromLandsat (Palisades,NY:NASASocioeconomicData andApplicationsCenter(SEDAC)) DemuzereMetal2014Mitigatingandadaptingtoclimate Codeavailability change:multi-functionalandmulti-scaleassessmentof greenurbaninfrastructureJ.Environ.Manage.146107–15 Scripts to process the raw temperature and rel- DosioA,MentaschiL,FischerEMandWyserK2018Extreme ◦ ◦ ative humidity observation are provided via heatwavesunder1.5 Cand2 CglobalwarmingEnviron. Res.Lett.13054006 https://github.com/jonasvdw/kampalasensors. EpsteinYandMoranDS2006Thermalcomfortandtheheat stressindicesInd.Health44388–98 ORCIDiDs FarrTGetal2007TheshuttleradartopographymissionRev. Geophys.45RG2004 FennerD,BechtelB,DemuzereM,KittnerJandMeierF2021 JVandeWalle https://orcid.org/0000-0002-7819- CrowdQC+—aquality-controlforcrowdsourced air-temperatureobservationsenablingworld-wideurban OBrousse https://orcid.org/0000-0002-7364- climateapplicationsFront.Environ.Sci.9553 FischerEM,OlesonKWandLawrenceDM2012Contrasting 710X urbanandruralheatstressresponsestoclimatechange CBrimicombe https://orcid.org/0000-0001-5550- Geophys.Res.Lett.39L03705 GreenH,BaileyJ,SchwarzL,VanosJ,EbiKandBenmarhniaT MDemuzere https://orcid.org/0000-0003-3237- 2019Impactofheatonmortalityandmorbidityinlowand middleincomecountries:areviewoftheepidemiological evidenceandconsiderationsforfutureresearchEnviron.Res. SLwasa https://orcid.org/0000-0003-4312-2836 17180–91 HMisiani https://orcid.org/0000-0002-5414-3125 GunawardenaK,WellsMandKershawT2017Utilisinggreenand GNsangi https://orcid.org/0000-0002-9324-106X bluespacetomitigateurbanheatislandintensitySci.Total Environ.584–51040–55 WThiery https://orcid.org/0000-0002-5183-6145 GuoYetal2014Globalvariationintheeffectsofambient BFZaitchik https://orcid.org/0000-0002-0698- temperatureonmortality:asystematicevaluation Epidemiology25781–9 NPMvanLipzig https://orcid.org/0000-0003- HalinskiRSandFeldtLS1970Theselectionofvariablesin multipleregressionanalysisJ.Educ.Meas.7151–7 2899-4046 HarringtonLJ,FrameDJ,FischerEM,HawkinsE,JoshiMand JonesCD2016Poorestcountriesexperienceearlier References anthropogenicemergenceofdailytemperatureextremes Environ.Res.Lett.11055007 AmouM,GyilbagA,DemelashTandXuY2021Heatwavesin HarringtonLJandOttoFEL2020Reconcilingtheorywiththe Kenya1987–2016:factsfromchirtshighresolutionsatellite realityofAfricanheatwavesNat.Clim.Change10796–8 remotelysensedandstationblendedtemperaturedataset HassAL,EllisKN,ReyesMasonL,HathawayJMandHoweDA Atmosphere1237 2016Heatandhumidityinthecity:neighborhoodheat 10 Environ. Res. Lett.17(2022)024004 JVandeWalleetal indexvariabilityinamid-sizedcityintheSoutheastern MarcotullioPJ,KeßlerCandFeketeBM2021Thefutureurban UnitedStatesInt.J.Environ.Res.PublicHealth13117 heat-wavechallengeinAfrica:exploratoryanalysisGlob. HemerijckxL-M,VanEmelenS,RymenantsJ,DavisJ, Environ.Change66102190 VerburgPH,LwasaSandVanRompaeyA2020Upscaling MastertonJMandRichardsonF1979Humidex:AMethodof householdsurveydatausingremotesensingtomap QuantifyingHumanDiscomfortduetoExcessiveHeatand socioeconomicgroupsinKampala,UgandaRemoteSens. Humidity(EnvironmentCanada,Atmospheric 123468 Environment)p45 HessJJetal2018Buildingresiliencetoclimatechange:pilot McGranahanGandSatterthwaiteD2014UrbanisationConcepts evaluationoftheimpactofIndia’sfirstheatactionplanon andTrendsvol220(JSTOR)(availableat:www.jstor. all-causemortalityJ.Environ.PublicHealth20187973519 org/stable/pdf/resrep01297.pdf?acceptTC=true&coverpage HintzMJ,LuederitzC,LangDJandvonWehrdenH2018Facing =false&addFooter=false) theheat:asystematicliteraturereviewexploringthe Medina-RamónMandSchwartzJ2007Temperature,temperature transferabilityofsolutionstocopewithurbanheatwaves extremesandmortality:astudyofacclimatisationandeffect UrbanClim.24714–27 modificationin50UScitiesOccup.Environ.Med.64827–33 HuangC,BarnettAG,WangX,VaneckovaP,FitzgeraldGand MeierF,FennerD,GrassmannT,OttoMandSchererD2017 TongS2011Projectingfutureheat-relatedmortalityunder Crowdsourcingairtemperaturefromcitizenweather climatechangescenarios:asystematicreviewEnviron. stationsforurbanclimateresearchUrbanClim.19170–91 HealthPerspect.1191681–90 MoraCetal2017GlobalriskofdeadlyheatNat.Clim.Change HubbartJ,LinkT,CampbellCandCobosD2005Evaluationofa 7501–6 low-costtemperaturemeasurementsystemfor MullerC,ChapmanL,JohnstonS,KiddC,IllingworthS, environmentalapplicationsHydrol.Process.191517–23 FoodyG,OvereemAandLeighR2015Crowdsourcingfor ImE-S,PalJSandEltahirEA2017Deadlyheatwavesprojected climateandatmosphericsciences:currentstatusandfuture inthedenselypopulatedagriculturalregionsofSouthAsia potentialInt.J.Climatol.353185–203 Sci.Adv.3e1603322 NagendraH,BaiX,BrondizioESandLwasaS2018Theurban IPCC2014ClimateChange2014PartA:GlobalandSectoral southandthepredicamentofglobalsustainabilityNat. Aspects(Cambridge:CambridgeUniversityPress)p1132 Sustain.1341–9 JonssonP,BennetC,EliassonIandSelinLindgrenE2004 NakamuraK1966CitytemperatureofNairobiJ.Geogr.75316–25 Suspendedparticulatematteranditsrelationstotheurban NastarM2020Messagesent,nowwhat?Acriticalanalysisofthe climateinDaresSalaam,TanzaniaAtmos.Environ. heatactionplaninAhmedabad,IndiaUrbanSci.453 384175–81 NdettoELandMatzarakisA2017Assessmentofhumanthermal KabanoP,LindleySandHarrisA2021Evidenceofurbanheat perceptioninthehot-humidclimateofDaresSalaam, islandimpactsonthevegetationgrowingseasonlengthina TanzaniaInt.J.Biometeorol.6169–85 tropicalcityLandscapeUrbanPlan.206103989 OkeTR1982TheenergeticbasisoftheurbanheatislandQ.J.R. KabumbuliRandKiwaziFW2009Participatoryplanning, Meteorol.Soc.1081–24 managementandalternativelivelihoodsforpoor OkeT1973CitysizeandtheurbanheatislandAtmos.Environ. wetland-dependentcommunitiesinKampala,UgandaAfr.J. 7769–79 Ecol.47154–60 OttoFEetal2020Challengestounderstandingextremeweather KampalaCapitalCityAuthorityandUgandaBureauofStatistics changesinlowerincomecountriesBull.Am.Meteorol.Soc. 2019StatisticalabstractforKampalaCity:reportprepared 101E1851–60 withsupportfromUgandaBureauofStatisticsTechnical OudinÅströmD,BertilFandJoacimR2011Heatwaveimpact Report pp1–163(availableat:www.kcca.go.ug/media/ onmorbidityandmortalityintheelderlypopulation:a docs/Statistical-Abstract-2019.pdf) reviewofrecentstudiesMaturitas6999–105 KnowltonKetal2014Developmentandimplementationof ParastatidisD,MitrakaZ,ChrysoulakisNandAbramsM2017 SouthAsia’sfirstheat-healthactionplaninAhmedabad Onlinegloballandsurfacetemperatureestimationfrom (Gujarat,India)Int.J.Environ.Res.PublicHealth LandsatRemoteSens.91208 113473–92 PotchterO,CohenP,LinT-PandMatzarakisA2018Outdoor KovatsRSandHajatS2008Heatstressandpublichealth:a humanthermalperceptioninvariousclimates:a criticalreviewAnnu.Rev.PublicHealth2941–55 comprehensivereviewofapproaches,methodsand LangeSetal2020Projectingexposuretoextremeclimateimpact quantificationSci.TotalEnviron.631390–406 eventsacrosssixeventcategoriesandthreespatialscales QuayleRandDoehringF1981HeatstressWeatherwise34120–4 Earth’sFuture8e2020EF001616 RichmondA,MyersIandNamuliH2018Urbaninformalityand LiX,StringerLCandDallimerM2021Thespatialandtemporal vulnerability:acasestudyinKampala,UgandaUrbanSci. characteristicsofurbanheatislandintensity:implications 222 forEastAfrica’surbandevelopmentClimate951 RothM2007Reviewofurbanclimateresearchin(sub)tropical LiY,SchubertS,KroppJPandRybskiD2020Ontheinfluenceof regionsInt.J.Climatol.271859–73 densityandmorphologyontheUrbanHeatIslandintensity RussoS,MarcheseAF,SillmannJandImméG2016Whenwill Nat.Commun.111–9 unusualheatwavesbecomenormalinawarmingAfrica? LiuZ,AndersonB,YanK,DongW,LiaoHandShiP2017Global Environ.Res.Lett.11054016 andregionalchangesinexposuretoextremeheatandthe SaeedF,SchleussnerC-FandAshfaqM2021Deadlyheatstressto relativecontributionsofclimateandpopulationchangeSci. becomecommonplaceacrossSouthAsiaalreadyat1.5 Cof Rep.71–9 globalwarmingGeophys.Res.Lett.48e2020GL091191 LwasaS,BuyanaK,KasaijaPandMutyabaJ2018Scenariosfor SalmondJAetal2016Healthandclimaterelatedecosystem adaptationandmitigationinurbanAfricaunder1.5 C servicesprovidedbystreettreesintheurbanenvironment globalwarmingCurr.Opin.Environ.Sustain.3052–58 Environ.Health1595–111 LwasaS,MugaggaF,WahabB,SimonD,ConnorsJandGriffithC ScottAAetal2017Temperatureandheatininformalsettlements 2014Urbanandperi-urbanagricultureandforestry: inNairobiPLoSOne121–17 transcendingpovertyalleviationtoclimatechange SinghR,ArrighiJ,JjembaE,StrachanK,SpiresMand mitigationandadaptationUrbanClim.792–106 KadihasanogluA2019HeatwaveguideforcitiesTechnical MalchaireJ,KampmannB,HavenithG,MehnertPand Report (Geneva:RedCrossRedCrescentClimateCentre) GebhardtH2000Criteriaforestimatingacceptable (availableat:www.climatecentre.org/downloads/files/IFRC exposuretimesinhotworkingenvironments:areviewInt. Geneva/RCCC%20Heatwave%20Guide%202019%20A4% Arch.Occup.Environ.Health73215–20 20RR%20ONLINE%20copy.pdf) 11 Environ. Res. Lett.17(2022)024004 JVandeWalleetal SouchCandSouchC1993Theeffectoftreesonsummertime VandeWalleJ2021Canlocalfieldworkhelptorepresent belowcanopyurbanclimates:acasestudyBloomington, intra-urbanvariabilityofcanopyparametersrelevantfor IndianaJ.Arboric.19303–12 tropicalAfricanclimatestudies?Theor.Appl.Climatol. StewartID2011RedefiningtheurbanheatislandPhDThesis 146457–74 UniversityofBritishColumbiap352(availableat: VandeWalleJ,ThieryW,BrousseO,SouverijnsN,DemuzereM https://circle.ubc.ca/handle/2429/38069) andvanLipzigNP2020Aconvection-permittingmodelfor StewartIDandOkeTR2012Localclimatezonesforurban theLakeVictoriaBasin:evaluationandinsightintothe temperaturestudiesBull.Am.Meteorol.Soc. mesoscaleversussynopticatmosphericdynamicsClim.Dyn. 931879–900 541779–99 StewartIandKennedyC2015Estimatinganthropogenicheat VanLeeuwenJM,SekeramayiT,MartellC,FeinbergMand releasefrommegacitiesICUC9–9thInt.Conf.onUrban Bowersox-DalyS2017AbaselineanalysisoftheKatanga ClimateHeldJointlywiththe12thSymp.ontheUrban slums:informingurbanpublicpolicyinKampala,Uganda Environment pp20–24(availableat:www.meteo. Afr.Popul.Stud.312 fr/icuc9/LongAbstracts/gd5-1-2961311_a.pdf) VenterZS,BrousseO,EsauIandMeierF2020Hyperlocal TahaH1997Urbanclimatesandheatislands:albedo, mappingofurbanairtemperatureusingremotesensingand evapotranspirationandanthropogenicheatEnergyBuild. crowdsourcedweatherdataRemoteSens.Environ. 2599–103 242111791 ThieryW,DavinEL,PanitzH-J,DemuzereM,LhermitteSand VermeirenK,VanRompaeyA,LoopmansM,SerwajjaEand vanLipzigN2015TheimpactoftheAfricanGreatLakeson MukwayaP2012UrbangrowthofKampala,Uganda: theregionalclimateJ.Clim.284061–85 patternanalysisandscenariodevelopmentLandscapeUrban ThieryW,DavinEL,SeneviratneSI,BedkaK,LhermitteSand Plan.106199–206 VanLipzigNP2016Hazardousthunderstorm VermeirenK,VanmaerckeM,BeckersJandVanRompaeyA2016 intensificationoverLakeVictoriaNat.Commun. ASSURE:amodelforthesimulationofurbanexpansion 71–7 andintra-urbansocialsegregationInt.J.Geogr.Inf.Sci. ThieryW,GudmundssonL,BedkaK,SemazziFHM, 302377–400 LhermitteS,WillemsP,vanLipzigNPMand WilbyRLetal2021Monitoringandmoderatingextremeindoor SeneviratneSI2017Earlywarningsofhazardous temperaturesinlow-incomeurbancommunities thunderstormsoverLakeVictoriaEnviron.Res.Lett. monitoringandmoderatingextremeindoortemperatures 12074012 inlow-incomeurbancommunitiesEnviron.Res.Lett TwinomuhangiR,SseviiriH,MulindeC,MukwayaPI, 16024033 NimusiimaAandKatoAM2021Perceptionsand WMO1986Guidelinesontheselectionofreference vulnerabilitytoclimatechangeamongtheurbanpoorin climatologicalstationsfromtheexistingclilmatological KampalaCity,UgandaReg.Environ.Change211–13 stationnetwork(availableat:https://library.wmo.int/ UgandaBureauofStatistics2014UgandaNationalPopulationand doc_num.php?explnum_id=9310) HousingCensus(Kampala:UgandaBureauofStatistics WoodhamsBJ,BarrettPA,MarshamJH,BirchCE,BainC, (UBOS))(availableat:https://unstats.un.org/unsd/demogra FletcherJ,HartleyAJandWebsterS2021Firstaircraft phic/sources/census/wphc/Uganda/UGA-2016-05-23.pdf) observationsofthelakeVictorialake–landbreeze UnitedNations2018WorldUrbanizationProspects2018: circulationfromtheHyVicpilotflightcampaign34thConf. Highlights(ST/ESA/SER.A/421)(availableat: onHurricanesandTropicalMeteorology(AMS)(availableat: https://population.un.org/wup/Publications/ https://ams.confex.com/ams/34HURR/meetingapp.cgi/ Files/WUP2018-Highlights.pdf) Paper/386612) UnitedNations2019WorldPopulationProspects2019:Highlights XuZandTongS2017Decomposetheassociationbetween (DepartmentofEconomicandSocialAffairs,Population heatwaveandmortality:whichtypeofheatwaveismore Division)(availableat:https://population.un.org/wpp/ detrimental?Environ.Res.156770–4 Publications/Files/WPP2019_Highlights.pdf) ZscheischlerJetal2018Futureclimateriskfromcompound vandeGiesenN,HutRandSelkerJ2014TheTrans-African eventsNat.Clim.Change8469–77 Hydro-MeteorologicalObservatory(TAHMO)WIREs ZscheischlerJetal2020Atypologyofcompoundweatherand Water 1341–8 climateeventsNat.Rev.EarthEnviron.1333–47 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Environmental Research Letters IOP Publishing

Lack of vegetation exacerbates exposure to dangerous heat in dense settlements in a tropical African city

Download PDF
12 pages

Loading next page...
 
/lp/iop-publishing/lack-of-vegetation-exacerbates-exposure-to-dangerous-heat-in-dense-L0cmRWLj33

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
IOP Publishing
Copyright
© 2022 The Author(s). Published by IOP Publishing Ltd
eISSN
1748-9326
DOI
10.1088/1748-9326/ac47c3
Publisher site
See Article on Publisher Site

Abstract

BothclimatechangeandrapidurbanizationaccelerateexposuretoheatinthecityofKampala, Uganda.Fromanetworkoflow-costtemperatureandhumiditysensors,operationalin2018–2019, wederivethedailymean,minimumandmaximumHumidexinordertoquantifyandexplain intra-urbanheatstressvariation.Thistemperature-humidityindexisshowntobeheterogeneously distributedoverthecity,withadailymeanintra-urbanHumidexIndexdeviationof1.2 Con average.Thelargestdifferencebetweenthecoolestandthewarmeststationoccursbetween16:00 and17:00localtime.Averagedoverthewholeobservationperiod,thisdailymaximumdifference ◦ ◦ is6.4 Cbetweenthewarmestandcooleststations,andreaches14.5 Conthemostextremeday. Thisheatstressheterogeneityalsotranslatestotheoccurrenceofextremeheat,showninother partsoftheworldtoputlocalpopulationsatriskofgreatdiscomfortorhealthdanger.Onestation inadensesettlementreportsadailymaximumHumidexIndexof >40 Cin68%ofthe observationdays,alevelwhichwasneverreachedatthenearbycampusoftheMakerereUniversity, andonlyafewtimesatthecityoutskirts.Largeintra-urbanheatstressdifferencesareexplainedby satelliteearthobservationproducts.NormalizedDifferenceVegetationIndexhasthehighest(75%) powertopredicttheintra-urbanvariationsindailymeanheatstress,butstrongcollinearityis foundwithothervariableslikeimpervioussurfacefractionandpopulationdensity.Ourresults haveimplicationsforurbanplanningontheonehand,highlightingtheimportanceofurban greening,andriskmanagementontheotherhand,recommendingtheuseofa temperature-humidityindexandaccountingforlargeintra-urbanheatstressvariationsand heat-pronedistrictsinurbanheatactionplansfortropicalhumidcities. 1.Introduction Schwartz2007,OudinÅströmetal2011,Fischeretal 2012, Mora et al 2017), raising serious concerns for Heat is a killer hazard with a global reach. Its expos- human health in a projected warmer future climate ure has been associated with both increased mortal- (Kovats and Hajat 2008, Huang et al 2011, Guo et al ity and morbidity worldwide (Medina-Ramón and 2014, Mora et al 2017). Former research on health ©2022TheAuthor(s). PublishedbyIOPPublishingLtd Environ. Res. Lett.17(2022)024004 JVandeWalleetal impactofextremeheatconcentratesonmid-latitude, intra-urbanvariations.LCZarethusexpectedtoalso high-income countries of low to medium popula- reflect heat stress variations in the city (Kabano et al tion density (Campbell et al 2018, Green et al 2019, 2021,VandeWalleetal2021),similartothefindings Otto et al 2020), thereby chronically underreporting in Nairobi (Kenya), concluding that informal settle- regions that are projected to actually experience the mentsareparticularlypronetoheatstress(Scottetal mostextremeheatinthefuture(Imetal2017,Mora 2017). etal2017,Nagendraetal2018,HarringtonandOtto However, observational studies investigating this 2020, Saeed et al 2021). For example, Africa is par- heterogeneity have been depreciated, because of the ticularly vulnerable to heat stress (IPCC 2014, Singh characteristic meteorological data scarcity in the et al 2019). A rapid increase in the intensities and region (Roth 2007). Six weather stations were set up frequenciesofheatwavesduringthepastdecadeshas inKampalaonlyrecently,thankstotheTrans-African beendemonstrated(Ceccherinietal2017,Amouetal Hydro-Meteorological Observatory (TAHMO, van 2021), while simulations project this trend to con- de Giesen et al 2014) project, collecting meteor- tinueuninterruptedintothefuture(Harringtonetal ological data from the synoptic station at Maker- 2016,Russoetal2016,Dosioetal2018).Forinstance, ere university and five instrument shelters placed in underahigheremissionscenario(SSP5-8.5),Africa’s open school gardens, in accordance with the offi- exposure to extreme heat is projected to be 7–269 cial World Meteorological Organization standards times larger than it has been historically (Liu et al (WMO1986).Despitethisgreatobservationaleffort, 2017,Asefi-Najafabadyetal2018). nostationsareplacedinmoredenselybuiltenviron- Extremeheatis furtheramplified incities, which mentswheremostofKampala’spopulationlives. areshowntobewarmerthantheirnaturalsurround- To better represent the variations of heat stress ings, because of reduced vegetated areas, increased throughout the city of Kampala, including densely release of anthropogenic heat, changes in surface populated areas, this study put in place an obser- albedo and trapped radiation within street canyons vational network of 45 low-cost iButton sensors. (Oke 1982). This urban heat island effect has also These sensors recorded near-surface air temperat- been demonstrated for Sub-Saharan African cities ure and relative humidity for three 42 d periods (Nakamura 1966, Jonsson et al 2004, Roth 2007, between August 2018 and April 2019 (Van de Walle Brousse et al 2020), experiencing rapid population et al 2021). From these measurements, the Humi- growth(McGranahanandSatterthwaite2014,United dex Index is computed, providing a good estimate Nations 2019) and extensive urbanization (Liu et al for feel-like temperature (Masterton and Richard- 2017, United Nations 2018, Marcotullio et al 2021). son 1979). High relative humidity decreases a per- As an example, Kampala, the capital city of Uganda, son’s evaporation ability and thereby the effective- is experiencing an uncontrolled urbanization, hav- ness of the body’s natural cooling system (Malchaire −1 ing the fourth highest growth rate (>4%yr ) of all et al 2000, Hass et al 2016). Particularly in hot and African cities (Richmond et al 2018, Kampala Cap- humid cities like Kampala, high values of the Humi- ital City Authority and Uganda Bureau of Statist- dex Index might cause dangerous health conditions. ics 2019). Like many fast-growing cities, Kampala is We therefore focus on extreme heat recorded at the expanding horizontally (Brousse et al 2019, Li et al different stations, and explain the observed patterns 2021),demonstratingspatialpatternsofurbansprawl basedonrelevantsatellite-derivedearthobservations. (Vermeiren et al 2016, Hemerijckx et al 2020) and For example, vegetation is known to generally play the formation of informal settlements or slums (Van a twofold role, decreasing temperature but enhan- Leeuwenetal2017,Lwasaetal2018,Richmondetal cing humidity by transpiration (Hass et al 2016). 2018). Results are discussed from two different perspect- Within the city of Kampala, both morpholo- ives:insightsinspatialheterogeneityofheatinKam- gicalandsocio-economicalcharacteristicslargelydif- pala and occurrences of heat above great discomfort fer, distinguishing wealthy districts characterized by thresholdsamongdifferenturbanenvironments. asphalted roads, modern houses and large gardens, frominformalsettlementscomposedofdenselybuilt 2.Methods shacksmadeofcorrugatedmetalsheetsthatareonly accessible via small alleys (Vermeiren et al 2012, 2.1.iButtonobservations Hemerijckxetal2020).Recently,Brousseetal(2019) The iButton sensor, a product of Maxim Integrated, classifiedtheseintra-urbanvariationsintoLocalCli- is a low-cost sensor containing a temperature and mateZones(LCZ,StewartandOke2012).Thisclassi- humidityloggingsystem(Hubbartetal2005).Witha ficationincludes7vegetatedand10builtclasses,each loggingfrequencyanddataaccuracyprogrammedat class exemplifying uniform surface cover, structure, 15 min and 11 bit respectively, each sensor can store material and human activity that span hundreds of 42 consecutive days of data. Afterwards, a manual meterstoseveralkilometersinhorizontalscale(Stew- download is required. To protect the sensors from art 2011). Importantly, LCZs are designed to reflect radiationandsplashwater,theyareshieldedbyafol- the thermal environments as a consequence of their dedthinlightreflectivefilm(figureS1availableonline 2 Environ. Res. Lett.17(2022)024004 JVandeWalleetal Figure1.LocationsofthedifferentsensorsthroughoutthecityofKampala,withtopviewoftheneighbourhoodsaroundthe stationsNamungoona(Ng),Makerere(Mk),Nakasero(Ns),Industrialarea(Ia),Nkeere(Nk)andBuziga(Bz).Asetofthree sensorsisinstalledperlocation(attheexactcentreofeachtopviewimage),reducingtheuncertaintyduetodifferentinstallation conditions.ExactlocationsofthesensorsarelistedintableS1.Eachtopviewimageis500 ×500m ,retrievedfromGoogleEarth imagery.Thelandsurfacetemperatureestimationforthecloud-freedayof27February2021isderivedfromLandsat8and MODISremotesensingproductsviaParastatidisetal(2017),andshowslargeintra-urbanvariationsofsurfacetemperatures. at stacks.iop.org/ERL/17/024004/mmedia), designed values. These periods do not cover a full year, yet we and produced by the Maryland Institute College monitored humidity and temperature for both dry of Art and Johns Hopkins University. The sensor- andwetseasons.Theintra-seasonalvariationsintem- containingshieldsarezip-tiedpreferablytoawooden peratureandhumidityaregenerallyratherlimitedin material,atabout2mheightandatashadedlocation the region (figure S3). On a longer timescale, warm (tableS1).Inthisstudywereusedthematerialofthe spellscanoccur,withthestartof2019asanexample. observational campaign by Scott et al (2017) held in The second of our observational periods covers that Nairobi,Kenya. period (figure S3). Due to a slightly different down- The resulting network of 45 sensors installed loading moment, measurement periods may slightly throughout the city of Kampala (figure 1) aimed at differ for locations located far from each other. In properlyrepresentingthecity’ssurfaceheterogeneity. addition, some data is missing for the second and Besides a good spatial coverage of both openly and third periods due to technical issues, especially low denselybuiltenvironments,thechoiceofsensorsloc- battery. At the end of the third period, 32 out of 45 ationalsoconsideredthesecurityofthesensorsfrom sensorsremainedactive. vandalism. Preferred locations thus included schools In addition, one sensor set was put indoor at the or houses of local acquaintances. At each location, a informal settlement of Acholi Quarters. The build- setofthreesensorswereinstalledclosetogether,redu- ing is a concrete block, no windows, an iron door cingtheuncertaintyofthetriple-sensor-meandueto andironroofsheets.Giventhefactthatitisonlyone different installation conditions such as attachment site, results are qualitatively described in the discus- material, shade fraction or ground cover. For each sionsection. location, the 15 min resolution triple-sensor-mean 2.2.Measurementquality information is reduced to the minimum and max- imum values per day. In addition, the daily mean is Themanufacturer’sevaluationoftheiButtonsreports a thermochron and hygrochron accuracy of 0.5 C computedastheaverageover24h.Threedownload- ing rounds provided data for 3 periods of 42 d each and 5% respectively. This is confirmed by calculat- ing the mean deviation per sensor triple, explained betweenAugust2018andApril2019(figureS2).For furtheranalyses,dailyminimumanddailymaximum and summarized in table S2. In addition, the triple-sensor-meanisevaluatedagainsttheMakerere valuesarecomputedastheaveragesoverall3 ×42d, stillreferredtoasdailyminimumanddailymaximum automatic weather station data, part of the TAHMO 3 Environ. Res. Lett.17(2022)024004 JVandeWalleetal network (van de Giesen et al 2014). With an overall vegetationcancounteracttheurbanheatislandeffect, ◦ ◦ ◦ temperature bias of 0.10 C, 0.11 C and 0.53 C for example by evaporative cooling (Oke 1982). The in the three periods, and relative humidity bias of normalized digital vegetation index (NDVI) ranging −2.50%, −2.66% and −1.43%, the iButton sensors from0to1,isusedasaproxyforthefractionofsur- tend to slightly overestimate the air temperature face covered by vegetation. This product is derived and underestimate humidity. This however varies from the Moderate Resolution Imaging Spectrora- throughout the day as observed nighttime temperat- diometer (MODIS) onboard the Terra satellite with ures by the iButton sensors are higher than the ones a horizontal resolution of 250 m. This satellite over- measured by the automatic weather station, while passesKampalaat10:30AMandPMlocaltime.Our observeddaytimetemperaturesarelower(figureS4). analysis uses the median value of two years, 2018– This results in an underestimation of the diurnal 2019, overlapping the observation period. No tem- temperature range. Nighttime relative humidity is poral variation is taken into account, assuming little underestimated by ∼4%, but daytime observations seasonal variation in this tropical area and keeping compare well to the Makerere automatic weather the focus to the spatial heterogeneity. Fourth, due station. to the high heat storage capacity of building mater- ials, densely built areas can largely influence local 2.3.Humidexindex temperatures (Oke 1973). As a proxy for this build- To better estimate the human-experienced heat, the ing density, impervious surface area (ISA) fraction Humidex Index (hereafter referred to as ‘Humi- is retrieved from the Global Man-made Impervious dex’,H)iscomputedevery15minfromobservations Surface(GMIS)dataset(DeColstounetal2017).For of both temperature (T in C) and relative humid- the target year 2010, the GMIS product analysed all ity (RH in %), following Masterton and Richardson cloud-free images from Landsat 5 and 7, inheriting (1979): thehighhorizontalresolutionof30m.Astrongcor- ( ) relation is expected, yet the abundance of bare soil 7.5 T 5 RH 237.7+T H(T,RH) =T + 6.112 10 −10 . oftenchallengessatelliteinstrumentstoproperlyrep- 9 100 resentISA(VandeWalleetal2021).Fifth,anthropo- (1) genicheat,mainlyfromdomesticandtransportation fueluse,isproducedinhighlypopulatedareas(Taha The resulting quantity increases non-linearly with 1997, Stewart and Kennedy 2015). The population both air temperature and relative humidity, and can densityofthegreaterKampalaregion,formallyavail- be understood as feel-like temperature in degrees ableperdistrictfortheyear2014(UgandaBureauof Celsius. Humidex values above 40 C lead to ‘great Statistics 2014, Hemerijckx et al 2020), is translated discomfort’, values exceeding 45 C are ‘dangerous’ to a 30 m resolution grid. Sixth, the MODIS instru- (MastertonandRichardson1979).Humidexinform- mentalsoprovidesdirectionalhemispherical(black- ationisreducedtotheminimumandmaximumval- sky) near-infrared albedo at 0.7–5.0 µm wavelength uesperday,aswellasthedailymeanwhichistheaver- at 500 m horizontal resolution, possibly distinguish- ageover24h. ingdifferentroofingtypeswithinthecityofKampala (Brest1987). 2.4.Explanatoryvariables IfthesevariablescanexplaintheobservedHumi- WeaimtoexplainspatialHumidexpatternsbycom- dexvariations,asimplestatisticalmodelcouldextra- paring them against potential explanatory factors, polate operational weather station data to the entire including distance to the lake, vegetation presence, city, providing heat stress information about hardly built-upfraction,surfaceelevation,populationdens- accessible locations such as informal settlements. ityandsurfacealbedo(figureS5).Thechoiceofeach Assuming linear behaviour, a multiple linear regres- ofthesesixfactorsisarguedbelow.First,locatednext sion technique is applied, expressing the Humidex to Lake Victoria, Kampala is affected by a daytime (H)intermsoftheexplanatoryvariablesx : lake breeze developing on a daily basis (Thiery et al 2015, 2016, 2017, Brousse et al 2020, Van de Walle et al 2020, Woodhams et al 2021). The distance of H = β + βx , (2) 0 i i thesensorstothelakeisthereforeusedtoaccountfor i=1 different onsets of cooling. Second, hills can optim- allybenefitfromcoolingwinds,lowerregionscannot. where the Ordinary Least Squares method estim- In addition, these low areas are typically wetlands, ates the best fitting coefficients β based on the providing opportunities for urban farming, but also observations.Initially,allN explanatoryvariablesare humidifying the air (Kabumbuli and Kiwazi 2009). included,butat-testdecidesontheeliminationofthe Therefore, a digital elevation model at 30 m hori- least significant variable. This backward elimination zontal resolution is retrieved from the Shuttle Radar continues iteratively until all explanatory variables Topography Mission (SRTM, Farr et al 2007) as a are significant at 95% confidence level. Ultimately, potentialexplanatoryvariable.Third,thepresenceof the remaining multiple linear regression model no 4 Environ. Res. Lett.17(2022)024004 JVandeWalleetal longer suffers from multicollinearity, ensuring its TheobservedHumidexheterogeneityisaresultof coefficientstobeoptimallystable(HalinskiandFeldt intra-urban temperature and relative humidity vari- 1970). ations.Thelatterareconsideredattimesofdailymin- imumandmaximumHumidex(figures2(a),(d)and (c), (f) respectively). Around sunrise, when Humi- 2.5.Localclimatezones dexreachesitsminimum,theairtemperaturealmost In addition to this quantitative approach, locations entirely determines the Humidex variations between are classified based on the building structure, land the sensor sites, with the urban air being clearly cover and human activity according to the LCZ clas- hotter and drier than at the outskirts (figures 2(a), sification scheme (Stewart and Oke 2012, Brousse (d) and (g)). Specific humidity between those sites etal2019,seetableS1).TheoutskirtsstationsBuloba is similar (not shown). Around noon, when Humi- (Bl), Kawanda (Kw), Bukerere (Bk) and Buziga (Bz) dex reaches its maximum, the situation is differ- are located in open low-rise environments (LCZ 6), ent. Then, the highest air temperatures are observed defined by small (3–10 m) buildings with abund- at Nateete (Nt), Nkeere (Nk) and Acholi Quarters ant plant cover (Stewart 2011). The campus of the (Aq,figure2(c)),allclassifiedaslightweightlow-rise Makerere University (Mk) is classified as open mid- LCZ 7. These locations also have the lowest relative rise class (LCZ 5), characterized by open arrange- humidity(figure2(f)).WhilethetemperatureatNaj- ment of 3–9 story buildings and abundant plant janankumbi(Nj,compactlow-riseLCZ3)isverysim- cover. The Industrial Area (Ia) station is located ilar to the temperatures at Nateete or Nkeere, it has in a large low-rise (LCZ 8) class, characterized by a clearly higher Humidex (figure 2(i)). This can be extensivepavedsurfacesbetweenlarge,lowbuildings, explained by Najjanankumbi’s high relative humid- oftenwithanindustrialorcommercialfunction.The ity compared to Natalee and Nkeere. We therefore Nakasero(Ns)stationislocatedincompactmid-rise needbothtemperatureandrelativehumiditytoprop- (LCZ 2) class, defined by buildings of 10–25 m sep- erly explain spatial variations in heat. Importantly, arated by narrow streets and inner courtyards, and neither Najjanankumbi’s air temperature or relative with few or no trees. In addition, we classify the sta- humidity is exceptional compared to other stations tions in Namungooma (Ng), Bwaise (Bw) and Naj- such as Nkeere or Makerere: the temperature distri- janankumbi (Nj) as compact low-rise (LCZ 3) and bution is similar to the one at Nkeere (figure S6(c)), Nateete(Nt),Nkeere(Nk)andAcholiQuarters(Aq) and relativehumidity valuesexceeding 70% occur as aslightweightlow-rise(LCZ7),oftencalledinformal frequent at Makerere (figure S6(f)). Instead, it is the settlements or slums. Both classes consist of small combination of both compound drivers that creates buildingstightlypackedalongnarrowstreetswithno high maximum Humidex values in Najjanankumbi or little vegetation. Typical for the latter class are the (figureS6(i),Zscheischleretal2018,2020). lightweight building materials (thatch, wood or cor- Relating the observed Humidex heterogeneity to rugatedmetal)andoftenformlessarrangementofthe the LCZ classification, the open low-rise (LCZ 6) buildings(Stewart2011). environments, together with the open mid-rise Makerere University campus (LCZ 5), report gen- erally lowest temperature, highest relative humidity 3.Results and lowest Humidex values (figure 2). These cool Measurements show clear differences in Humidex environments contrast with the warm compact and values at the different sensor locations (figures 2(g)– lightweight low-rise classes (LCZ 3 and LCZ 7), as (i)). For example, the average Humidex varies well as the compact mid-rise central business centre ◦ ◦ between 30.6 C and 32.1 C for the urban stations, (LCZ 2). Kampala’s industrial area (Ia, LCZ 8) is exceptfortheMakerere(Mk)station,locatednearthe rather warm at night, but shows relatively mild tem- city centre, with a substantially lower average Humi- peraturesduringdaytime.Especiallypoorlyvegetated dex (29.0 C). Also the city’s outskirts, represented andcompactlybuiltneighbourhoodsinKampalaare by Buloba (Bl), Bukerere (Bk) and Kawanda (Kw) thus more prone to heat stress than the outskirts or ◦ ◦ stations, are cooler (29.3 C–29.9 C, figure 2(h)). Makererestation. At night (figure 2(g)), both the city outskirts and Intra-urban differences are especially relevant Makerere are cool, in contrast with high Humidex whenconsideringextremeheat(figure3).Concretely, values observed nearby central stations. Particularly Makerere (Mk) has no single day with the Humidex theindustrialarea(Ia),Nakasero(Ns),Namungoona exceeding the ‘great discomfort’ threshold of 40 C, (Ng)andNkeere(Nk)experiencewarmnights,with itoccursin2%–16%oftheobserveddaysintheout- differences up to 2.3 C compared to Makerere. skirtsstationsBukerere(Bk),Kawanda(Kw),Buloba The intra-urban heterogeneity is most pronounced (Bl) and Buziga (Bz), and 14%–67% in the stations when comparing daily maximum Humidex values locatedindenselybuiltareas,inparticularinNkeere (figure 2(i)), ranging from 34.1 C at Makerere to (Nk),AcholiQuarters(Aq)andNajjanankumbi(Nj) 41.6 CatNajjanankumbi. (in50%–67%oftheobserveddays).Lookingatdays 5 Environ. Res. Lett.17(2022)024004 JVandeWalleetal Figure2.Observedtemperatureandrelativehumidityatalllocations,aswellastheHumidexderivedfollowingequation(1).The leftandrightcolumnsshowobservationsatdailyminimumandmaximumHumidexrespectively,whilethemiddlecolumngives theHumidexaveragedoverthethreeobservationperiods(3 ×42d).ThegreycontourlineindicatestheKampalaurbanextent, definedbyISAfractionabove0.1(figureS5).DifferentmarkersymbolscorrespondtothedifferentLCZclassesobtainedfrom Brousseetal(2019)andVandeWalleetal(2021). with maximum Humidex exceeding 45 C, the ‘dan- correlations are found between Humidex and the gerous’threshold,Bwaise(Bw),Najjanankumbi(Nj) proximity to the lake or elevation. Importantly, the and to a lesser extent Nkeere (Nk) stand out, with explanatoryvariablesarenotindependentfromeach occurrences of 17%, 12% and 4% of the observed other, with high ISA fractions prohibiting abundant days,respectively. vegetation(correlationof −0.93)whilealsoimplying Not only for daily maximum Humidex, also lower near-infrared albedo due to the strongly mod- for daily minimum or average Humidex, Nkeere ified land cover (correlation of −0.82, figure S10). (Nk), Acholi Quarters (Aq) and Namungoona (Ng) Also population density is not independent from show highest exceedance frequencies of all Humi- ISA fraction, NDVI or near-infrared albedo, with dexthresholds(figureS7).Whenalsoaccountingfor correlations of 0.71, −0.78 and −0.72, respectively. the duration of exceedance by considering the hours The elevation shows a moderate correlation of 0.51 above a certain threshold (figure S8, top), or for the with NDVI, probably related to Kampala’s vegetated heat intensity by computing the mean heat-degree- hills. hours(figureS8,bottom),thedenselybuiltenviron- Due to this collinearity between the explanat- mentsoftenexperienceextremeheat(1%–10%ofthe oryvariables,thestepwisebackwardeliminationpro- observedtime),especiallyinNajjanankumbi(Nj). cedure only retains NDVI as explanatory variable The intra-urban heterogeneous Humidex values in the linear regression model for minimum, mean correlates with six proposed explanatory variables and maximum Humidex (figure 4). With R =0.79, (figureS9).Strongestcorrelationsarefoundbetween the NDVI has a high explanatory power for min- Humidex and NDVI as well as ISA fraction. Moder- imum (early morning) Humidex, meaning that 79% ate negative correlations of −0.5 to −0.65 are found of the the intra-urban Humidex variability can be between Humidex and near-infrared albedo, while explained.Thisexplanatorypowerissimilarforaver- Humidex and population density are positively cor- age(75%),butlowerformaximum(midday)Humi- related, with values between 0.5 and 0.76. Smaller dex(52%). 6 Environ. Res. Lett.17(2022)024004 JVandeWalleetal Figure3.Exposuretimetoheatstressextremesforeachlocation.Timeisgiveninpercentageofobserveddaysoutof3 ×42dat whichthedailymaximumHumidexexceedsacertainthreshold,definedbythevaluesontheabscissa.Typicalthresholdsare ◦ ◦ ◦ indicated,relatedto‘somediscomfort’ifH >30 C,to‘greatdiscomfort’ifH >40 Candto‘dangerous’ifH >45 C. max max max Figure4.(b)DependencyoftheHumidexaveragedovertheperiodof3 ×42donvegetationfraction(NDVI)foreachlocation. Samefordependenciesofdailyminimum(a)andmaximum(c)Humidex.ColourscorrespondtotheLCZclassesassignedbased ontheirGoogleEarthtop-viewimages(seefigure1).Linearregressionresultsafterastepwisebackwardeliminationprocedureof otherexplanatoryvariablesisshownbythefullbluelineandthecorrespondingequation,whiledashedlinesdefinethe95% 2 2 confidencebands.TheR valuesprovidetheexplanatorypoweroftheregressionmodels,withR adjustingforthenumberof adj predictors. Figure5.Extrapolationofdailyminimum(a),mean(b)andmaximum(c)Humidexbasedontheregressionmodelsfrom figure4.Inpractice,theHumidexisderivedfromtheexplanatoryvariableNDVIavailableat250mresolution(figureS5), supplementedbyandmultipliedwithtwocoefficients β and β (equation(2))resultingfromthelinearregressionprocedure. 0 1 The explanatory power of NDVI allows us to 4.Discussion applytheregressionmodelbyextrapolatingthepoint observations of minimum, mean and maximum Withaverageintra-urbandifferencesof1.2 C,andan Humidex to the greater Kampala region (figure 5). afternoondifferenceof6.4 Conaverage,theHumi- Thismapprovidesinformationonthespatialhetero- dex Index is heterogeneously distributed over the geneityofheatstressinKampala. city of Kampala. These large intra-urban differences 7 Environ. Res. Lett.17(2022)024004 JVandeWalleetal are also reflected in the number of days exposed to open low-rise (LCZ 6) and open mid-rise (LCZ 5) extremeheat.Atsomelocations,dailymaximumheat classes. Overall, sparsely built and highly vegetated stress exceeds the great discomfort level, defined by areasthusexperiencesubstantiallylowerheat,withan ◦ ◦ Humidex >40 C, for more than 50% of the 3 × observedmeandifferenceof6.4 Cintheafternoon. 42 observation days. In comparison, for the same This result implies that greening the city could mit- period, this level was never reached at the Maker- igate urban heat (Bowler et al 2010, Demuzere et al ere station and only a few times in the city outskirts. 2014, Gunawardena et al 2017). Yet, a more detailed Moreover, we identified the regions in Kampala that investigationisneededtotheoveralleffectsofdiffer- are most prone to heat, pointed to non-vegetated, enttypesofvegetationonhumanwell-being,includ- densely built environments using linear regression ing the effects on local climate, air quality and aes- and extrapolated the result to the greater Kampala thetics (Salmond et al 2016). In fact, the Kampala region.Theresultingmapcouldcomplementremote City Council Authority (KCCA) announced plans to sensingproductsforlandsurfacetemperature,adding plant0.5milliontreesinKampalaaspartofitsclimate information about 2 m temperature and relative action strategy, which is developed with stakehold- humidity combined in the Humidex Index. Such ers. Concretely, the strategy has embarked on tak- map can guide anticipatory action plans that help ing stock of the trees in the city coupled with map- reduce the impact of heatwaves, by providing con- ping of natural assets in the city to form the basis crete information about heat-prone areas. With spe- for implementation of the climate strategy. This cli- cialattentiontothoseheat-proneareas,aheataction mateactionstrategyischallenging,especiallybecause plan commits to public awareness about heat risks most available land is in Kampala’s residential areas (Singh et al 2019), which has been demonstrated to where urban tree canopies are already evident, while be successful in reducing heat-related mortality in we showed that the need for cooling is most urgent Ahmedabad, India (Knowlton et al 2014, Hess et al in densely built environments and informal settle- 2018, Nastar 2020). In addition, a heat action plan ments. Yet, Lwasa et al (2014) claimed the potential also accounts for the vulnerability of urban dwell- forexpansionoftreecanopycoverinthecity’sdensely ers.Ingeneral,denselybuiltandinformalsettlements built and most vulnerable areas as well. For this, housealargepartofthepopulationbelongingtothe initiatives from local actors and inhabitants should lower socioeconomic status with income and liveli- be strongly supported, which can only be achieved hoodinsecurity,makingthemparticularlyvulnerable by properly informing the inhabitants (Hintz et al (Vermeiren et al 2012, Lwasa et al 2018, Hemerijckx 2018), followed by incentive-based mechanisms that et al 2020, Twinomuhangi et al 2021). An import- allow the planting, nurturing and taking care of the ant factor increasing their vulnerability is the hous- trees in dense settlements by the developers. As an inginfrastructure,notofferinganyprotectiontoheat. example,somelocalpoliticiansandinfluencerschal- Asatestcase,wecollectedheatobservationsinsidea lengedUgandansvia(social)mediatocommemorate building in the informal settlement of Acholi Quar- everymarriage,death,birthandgraduationbyplant- ters during the three periods of the observational ingapersonaltree. campaign.Insteadofofferingprotectionagainstheat, Fiveimportantchallengeshavetobeaddressedby the house seems to act like a heat trap for even- future research. First, besides Humidex, many other ingandnighttimeheat,timeswhenpeopleareliving indices can describe heat as a combination of mul- inside.Onlyheatbeforeandatnoonisslightlylower tiplefactorsinfluencingheatstress.Toexplainhuman thenoutsideobservations(figureS11).Arecentstudy discomfort, many factors play a role, including air in Ghana investigated indoor temperatures and con- temperatureandhumidity(EpsteinandMoran2006, cludedlargeeffectsofbuildingmaterials(Wilbyetal Barnett et al 2010, Fischer et al 2012, Lange et al 2021). 2020),butalsowind,radiation,physiologicalfactors, A second implication concerns urban adaptation physical activity and clothing (Quayle and Doehring planning, and follows from our finding that Kam- 1981, Roth 2007, Potchter et al 2018). This study pala’s intra-urban varying heat is strongly correl- includestemperatureandrelativehumidityonly,with ated with NDVI, explaining up to 77% of the intra- differences between similar heat measures shown to urban variability in daily mean Humidex. Despite be small (Barnett et al 2010). Second, to explain the their elimination in the regression analysis due to observed Humidex values, vegetation is an import- clearcorrelationswithNDVI,otherexplanatoryvari- ant factor with a twofold role. On the one hand, ables,particularlyISAfraction,mightalsobeimport- enhancedtranspirationbyvegetationimpliesairtem- ant. From a Local Climate Zone perspective (Stew- perature cooling, tending to reduce the Humidex. art and Oke 2012), the warmest stations were found Ontheotherhand,enhancedtranspirationincreases in compact LCZ classes, characterized by densely the humidity which contributes to a higher Humi- built environments with little or no vegetation. In dex (Hass et al 2016). By showing a strong negative particular, they include compact mid-rise (LCZ 2), correlation between Humidex and vegetation frac- compact low-rise (LCZ 3) and lightweight (LCZ 7) tion (NDVI), we conclude that the first role dom- classes. Cooler environments in Kampala include inates. Concretely, adding vegetation might increase 8 Environ. Res. Lett.17(2022)024004 JVandeWalleetal the relative humidity, the reduction in temperat- In some areas, heat is experienced as ‘great discom- ure is superior. Yet this study did not quantitatively fort’below/abovetheproposed40 Cthreshold.Sur- investigate the interplay between temperature and veys in Dar es Salaam (Tanzania) suggested that the humidity in detail. A study in Indiana, US, high- comfort range was well above the one defined in lighted the importance of different vegetation types temperate climates (Ndetto and Matzarakis 2017). and structures to unravel this interplay (Souch and Whileconcreteinformationaboutapplicablelevels/- Souch 1993). Individual tree species or open grass thresholds for Kampala is lacking, this study provi- areas have the largest cooling effect, street trees the sionally explored different thresholds for the Humi- ◦ ◦ smallest. Also in Kampala, different vegetation types dexIndex,rangingfrom30 Cto50 C.Fifth,though could be obtained or derived from detailed land- the explanatory power of the regression model is covermaps.Evenifsuchland-covermapswouldhelp high(75%),fourfactorscouldleadtoimprovements the interpretation of the results by further linking of the model. First, it would definitely benefit from heat to different vegetation types, our study already moreobservations,perhapssupplementedbyremote integrates the level of greenness and vegetation by sensing data. Second, this study explored only six using NDVI and ISA variables. Third, our observa- explanatory variables, but more could be added, for tional iButton network was only active for a short 3 exampleprovidinginformationaboutbuildingmor- × 42 d period, during a warm spell. Results might phology or materials. Though this information is therefore slightly overestimate the multi-year situ- implicitlyalreadyincludedinLCZclasses,highresol- ation. In addition, our network was not complete ution material maps will possibly appear in the near to cover all neighbourhoods. Moreover, paper-made future. Third, future research should investigate the shields might not be sufficient to perfectly protect robustness of the model when adding new observa- the sensor from direct sunlight. However, the role of tions. Fourth, we found strong collinearity between irradiation of the sensor shields is not investigated different explanatory factors, challenging the causal- since the time of the day at which the sensor is sun- ity between vegetation fraction and Humidex Index lit/shaded was not inventorized. Deviations between inourlinearregressionmodel. the sensor triple observations are small, suggesting that the effect of irradiation is also small. Despite 5.Conclusion thegoodperformanceofoursensors,westillrecom- mend using a pair or triplet of iButtons per site, Fromanetworkoflow-costtemperatureandhumid- regardless of the reduced number of observational ity sensors, we compute the Humidex Index, quanti- sites. Besides our iButton network, weather station fying the heat stress throughout the city. Daily min- observationsareincreasinglyconductedintheregion imum,mean,maximumaswellasextremeheatstress via the TAHMO project (van de Giesen et al 2014), are heterogeneously distributed over the city, with yetthemeteorologicalnetworkisstillsparse.Besides poorly vegetated and densely built-up environments TAHMO, the availability of crowd-sourced weather being the most heat-prone areas. Their inhabitants, data has grown worldwide in recent years, offering generally vulnerable people due to their socioeco- possibilities for high-resolution observational urban nomic status, are often exposed to great discom- heat studies (Muller et al 2015, Chapman et al 2017, fort or even dangerous heat. Future research should Meier et al 2017, Venter et al 2020, Fenner et al bridge the gap to indoor heat and its dependence on 2021). Unfortunately, no studies have collected and house structure and building materials. Two recom- analysed such data in tropical Africa. Fourth, there mendations follow from this study: to mitigate heat is no common definition of heatwaves (Hintz et al stress,urbangreeningshouldbeconsideredinurban 2018). This study investigated different definitions planning strategies, and urban heat action plans including total days/nights or hours with extreme should account for the large intra-urban heat stress heat and heat degree hours. Similar to Europe, the variations. US and Australia, a region-specific heatwave defini- tion should be derived from mortality and morbid- ity studies (Kovats and Hajat 2008, Guo et al 2014, Dataavailabilitystatement Mora et al 2017, Xu and Tong 2017, Li et al 2020). Yet such studies are currently completely lacking The data that support the findings of this study for Africa (Harrington and Otto 2020). In addition, are openly available at the following URL/DOI: given an appropriate metric, health impact research http://doi.org/10.5281/zenodo.5105570. should also question whether certain thresholds are Other products are publicly available online: the meaningful for different regions of interest based TAHMOautomaticweatherstationdataatMakerere on more metabolic information on people living in University (https://tahmo.org/climate-data/), SRTM these regions. In this study we provided the gen- Digital elevation model (https://gee.stac.cloud/6FKB eral (dis)comfort levels accompanying the Humidex uAFUoXyYMZCUkttkJ5VuS8cQd?t = bands), Index. Yet, experienced heat is expected to depend MODIS MYD13Q1 product for NDVI (https://mo on the region of the world (Potchter et al 2018). dis.gsfc.nasa.gov/data/dataprod/mod13.php), GMIS 9 Environ. Res. Lett.17(2022)024004 JVandeWalleetal built-up fraction (https://sedac.ciesin.columbia.edu/ Asefi-NajafabadyS,VandecarKL,SeimonA,LawrencePand LawrenceD2018Climatechange,populationandpoverty: data/set/ulandsat-gmis-v1), Kampala population vulnerabilityandexposuretoheatstressincountries density (https://unstats.un.org/unsd/demographic/ borderingthegreatlakesofAfricaClim.Change148561–73 sources/census/wphc/Uganda/UGA-2016-05-23.pdf) BarnettAG,TongSandClementsAC2010Whatmeasureof and MODIS MCD43A3 v6 albedo product (https:// temperatureisthebestpredictorofmortality?Environ.Res. 110604–11 modis.gsfc.nasa.gov/data/dataprod/mod43.php). In BowlerDE,Buyung-AliL,KnightTMandPullinAS2010Urban addition,thelandsurfacetemperatureisderivedfrom greeningtocooltownsandcities:asystematicreviewofthe Landsat 8 and MODIS, and can be retrieved online empiricalevidenceLandscapeUrbanPlan.97147–55 via:http://rslab.gr/downloads_LandsatLST.html BrestCL1987Seasonalalbedoofanurban/rurallandscapefrom satelliteobservationsJ.Appl.Meteorol.Climatol.261169–87 BrousseOetal2019UsinglocalclimatezonesinSub-Saharan Acknowledgments AfricatotackleurbanhealthissuesUrbanClim.27227–42 BrousseO,WoutersH,DemuzereM,ThieryW,VandeWalleJ andvanLipzigNP2020Thelocalclimateimpactofan ThisworkwasfinanciallysupportedbyInternalSpe- Africancityduringclear-skyconditions—implicationsof cialResearchFund(BOF-479C1ProjectC14/17/053) therecenturbanizationinKampala(Uganda)Int.J. and the REACT project. M Demuzere is suppor- Climatol.404586–608 ted by the ENLIGHT project, funded by the Ger- CampbellS,RemenyiTA,WhiteCJandJohnstonFH2018 Heatwaveandhealthimpactresearch:aglobalreviewHealth man Research Foundation (DFG) under Grant No. Place53210–8 437467569. The authors thank Daniel S Ddumba CeccheriniG,RussoS,AmeztoyI,FrancescoMarcheseAand (Makerere University) and Sadam Y Kiwanuka Carmona-MorenoC2017HeatwavesinAfrica1981–2015, (KCCA) for their support during the field work, and observationsandreanalysisNat.HazardsEarthSyst.Sci. 17115–25 local people for taking care of the sensors. OB was ChapmanL,BellCandBellS2017Canthecrowdsourcingdata financed by the Remote Sensing for Epidemiology paradigmtakeatmosphericsciencetoanewlevel?Acase in Sub-Saharan African CiTies (REACT http://react. studyoftheurbanheatislandofLondonquantifiedusing ulb.be/)project,fundedbytheSTEREO-IIIprogram NetatmoweatherstationsInt.J.Climatol.373597–605 DeColstounECB,HuangC,WangP,TiltonJC,PhillipsJ, of the Belgian Science Policy (BELSPO, SR/00/337) NiemczuraS,LingP-YandWolfeR2017Documentationfor andtheWellcomeTrustHEROICproject(GrantNo. theGlobalMan-madeImperviousSurface(GMIS)Dataset 216035/Z/19/Z). FromLandsat (Palisades,NY:NASASocioeconomicData andApplicationsCenter(SEDAC)) DemuzereMetal2014Mitigatingandadaptingtoclimate Codeavailability change:multi-functionalandmulti-scaleassessmentof greenurbaninfrastructureJ.Environ.Manage.146107–15 Scripts to process the raw temperature and rel- DosioA,MentaschiL,FischerEMandWyserK2018Extreme ◦ ◦ ative humidity observation are provided via heatwavesunder1.5 Cand2 CglobalwarmingEnviron. Res.Lett.13054006 https://github.com/jonasvdw/kampalasensors. EpsteinYandMoranDS2006Thermalcomfortandtheheat stressindicesInd.Health44388–98 ORCIDiDs FarrTGetal2007TheshuttleradartopographymissionRev. Geophys.45RG2004 FennerD,BechtelB,DemuzereM,KittnerJandMeierF2021 JVandeWalle https://orcid.org/0000-0002-7819- CrowdQC+—aquality-controlforcrowdsourced air-temperatureobservationsenablingworld-wideurban OBrousse https://orcid.org/0000-0002-7364- climateapplicationsFront.Environ.Sci.9553 FischerEM,OlesonKWandLawrenceDM2012Contrasting 710X urbanandruralheatstressresponsestoclimatechange CBrimicombe https://orcid.org/0000-0001-5550- Geophys.Res.Lett.39L03705 GreenH,BaileyJ,SchwarzL,VanosJ,EbiKandBenmarhniaT MDemuzere https://orcid.org/0000-0003-3237- 2019Impactofheatonmortalityandmorbidityinlowand middleincomecountries:areviewoftheepidemiological evidenceandconsiderationsforfutureresearchEnviron.Res. SLwasa https://orcid.org/0000-0003-4312-2836 17180–91 HMisiani https://orcid.org/0000-0002-5414-3125 GunawardenaK,WellsMandKershawT2017Utilisinggreenand GNsangi https://orcid.org/0000-0002-9324-106X bluespacetomitigateurbanheatislandintensitySci.Total Environ.584–51040–55 WThiery https://orcid.org/0000-0002-5183-6145 GuoYetal2014Globalvariationintheeffectsofambient BFZaitchik https://orcid.org/0000-0002-0698- temperatureonmortality:asystematicevaluation Epidemiology25781–9 NPMvanLipzig https://orcid.org/0000-0003- HalinskiRSandFeldtLS1970Theselectionofvariablesin multipleregressionanalysisJ.Educ.Meas.7151–7 2899-4046 HarringtonLJ,FrameDJ,FischerEM,HawkinsE,JoshiMand JonesCD2016Poorestcountriesexperienceearlier References anthropogenicemergenceofdailytemperatureextremes Environ.Res.Lett.11055007 AmouM,GyilbagA,DemelashTandXuY2021Heatwavesin HarringtonLJandOttoFEL2020Reconcilingtheorywiththe Kenya1987–2016:factsfromchirtshighresolutionsatellite realityofAfricanheatwavesNat.Clim.Change10796–8 remotelysensedandstationblendedtemperaturedataset HassAL,EllisKN,ReyesMasonL,HathawayJMandHoweDA Atmosphere1237 2016Heatandhumidityinthecity:neighborhoodheat 10 Environ. Res. Lett.17(2022)024004 JVandeWalleetal indexvariabilityinamid-sizedcityintheSoutheastern MarcotullioPJ,KeßlerCandFeketeBM2021Thefutureurban UnitedStatesInt.J.Environ.Res.PublicHealth13117 heat-wavechallengeinAfrica:exploratoryanalysisGlob. HemerijckxL-M,VanEmelenS,RymenantsJ,DavisJ, Environ.Change66102190 VerburgPH,LwasaSandVanRompaeyA2020Upscaling MastertonJMandRichardsonF1979Humidex:AMethodof householdsurveydatausingremotesensingtomap QuantifyingHumanDiscomfortduetoExcessiveHeatand socioeconomicgroupsinKampala,UgandaRemoteSens. Humidity(EnvironmentCanada,Atmospheric 123468 Environment)p45 HessJJetal2018Buildingresiliencetoclimatechange:pilot McGranahanGandSatterthwaiteD2014UrbanisationConcepts evaluationoftheimpactofIndia’sfirstheatactionplanon andTrendsvol220(JSTOR)(availableat:www.jstor. all-causemortalityJ.Environ.PublicHealth20187973519 org/stable/pdf/resrep01297.pdf?acceptTC=true&coverpage HintzMJ,LuederitzC,LangDJandvonWehrdenH2018Facing =false&addFooter=false) theheat:asystematicliteraturereviewexploringthe Medina-RamónMandSchwartzJ2007Temperature,temperature transferabilityofsolutionstocopewithurbanheatwaves extremesandmortality:astudyofacclimatisationandeffect UrbanClim.24714–27 modificationin50UScitiesOccup.Environ.Med.64827–33 HuangC,BarnettAG,WangX,VaneckovaP,FitzgeraldGand MeierF,FennerD,GrassmannT,OttoMandSchererD2017 TongS2011Projectingfutureheat-relatedmortalityunder Crowdsourcingairtemperaturefromcitizenweather climatechangescenarios:asystematicreviewEnviron. stationsforurbanclimateresearchUrbanClim.19170–91 HealthPerspect.1191681–90 MoraCetal2017GlobalriskofdeadlyheatNat.Clim.Change HubbartJ,LinkT,CampbellCandCobosD2005Evaluationofa 7501–6 low-costtemperaturemeasurementsystemfor MullerC,ChapmanL,JohnstonS,KiddC,IllingworthS, environmentalapplicationsHydrol.Process.191517–23 FoodyG,OvereemAandLeighR2015Crowdsourcingfor ImE-S,PalJSandEltahirEA2017Deadlyheatwavesprojected climateandatmosphericsciences:currentstatusandfuture inthedenselypopulatedagriculturalregionsofSouthAsia potentialInt.J.Climatol.353185–203 Sci.Adv.3e1603322 NagendraH,BaiX,BrondizioESandLwasaS2018Theurban IPCC2014ClimateChange2014PartA:GlobalandSectoral southandthepredicamentofglobalsustainabilityNat. Aspects(Cambridge:CambridgeUniversityPress)p1132 Sustain.1341–9 JonssonP,BennetC,EliassonIandSelinLindgrenE2004 NakamuraK1966CitytemperatureofNairobiJ.Geogr.75316–25 Suspendedparticulatematteranditsrelationstotheurban NastarM2020Messagesent,nowwhat?Acriticalanalysisofthe climateinDaresSalaam,TanzaniaAtmos.Environ. heatactionplaninAhmedabad,IndiaUrbanSci.453 384175–81 NdettoELandMatzarakisA2017Assessmentofhumanthermal KabanoP,LindleySandHarrisA2021Evidenceofurbanheat perceptioninthehot-humidclimateofDaresSalaam, islandimpactsonthevegetationgrowingseasonlengthina TanzaniaInt.J.Biometeorol.6169–85 tropicalcityLandscapeUrbanPlan.206103989 OkeTR1982TheenergeticbasisoftheurbanheatislandQ.J.R. KabumbuliRandKiwaziFW2009Participatoryplanning, Meteorol.Soc.1081–24 managementandalternativelivelihoodsforpoor OkeT1973CitysizeandtheurbanheatislandAtmos.Environ. wetland-dependentcommunitiesinKampala,UgandaAfr.J. 7769–79 Ecol.47154–60 OttoFEetal2020Challengestounderstandingextremeweather KampalaCapitalCityAuthorityandUgandaBureauofStatistics changesinlowerincomecountriesBull.Am.Meteorol.Soc. 2019StatisticalabstractforKampalaCity:reportprepared 101E1851–60 withsupportfromUgandaBureauofStatisticsTechnical OudinÅströmD,BertilFandJoacimR2011Heatwaveimpact Report pp1–163(availableat:www.kcca.go.ug/media/ onmorbidityandmortalityintheelderlypopulation:a docs/Statistical-Abstract-2019.pdf) reviewofrecentstudiesMaturitas6999–105 KnowltonKetal2014Developmentandimplementationof ParastatidisD,MitrakaZ,ChrysoulakisNandAbramsM2017 SouthAsia’sfirstheat-healthactionplaninAhmedabad Onlinegloballandsurfacetemperatureestimationfrom (Gujarat,India)Int.J.Environ.Res.PublicHealth LandsatRemoteSens.91208 113473–92 PotchterO,CohenP,LinT-PandMatzarakisA2018Outdoor KovatsRSandHajatS2008Heatstressandpublichealth:a humanthermalperceptioninvariousclimates:a criticalreviewAnnu.Rev.PublicHealth2941–55 comprehensivereviewofapproaches,methodsand LangeSetal2020Projectingexposuretoextremeclimateimpact quantificationSci.TotalEnviron.631390–406 eventsacrosssixeventcategoriesandthreespatialscales QuayleRandDoehringF1981HeatstressWeatherwise34120–4 Earth’sFuture8e2020EF001616 RichmondA,MyersIandNamuliH2018Urbaninformalityand LiX,StringerLCandDallimerM2021Thespatialandtemporal vulnerability:acasestudyinKampala,UgandaUrbanSci. characteristicsofurbanheatislandintensity:implications 222 forEastAfrica’surbandevelopmentClimate951 RothM2007Reviewofurbanclimateresearchin(sub)tropical LiY,SchubertS,KroppJPandRybskiD2020Ontheinfluenceof regionsInt.J.Climatol.271859–73 densityandmorphologyontheUrbanHeatIslandintensity RussoS,MarcheseAF,SillmannJandImméG2016Whenwill Nat.Commun.111–9 unusualheatwavesbecomenormalinawarmingAfrica? LiuZ,AndersonB,YanK,DongW,LiaoHandShiP2017Global Environ.Res.Lett.11054016 andregionalchangesinexposuretoextremeheatandthe SaeedF,SchleussnerC-FandAshfaqM2021Deadlyheatstressto relativecontributionsofclimateandpopulationchangeSci. becomecommonplaceacrossSouthAsiaalreadyat1.5 Cof Rep.71–9 globalwarmingGeophys.Res.Lett.48e2020GL091191 LwasaS,BuyanaK,KasaijaPandMutyabaJ2018Scenariosfor SalmondJAetal2016Healthandclimaterelatedecosystem adaptationandmitigationinurbanAfricaunder1.5 C servicesprovidedbystreettreesintheurbanenvironment globalwarmingCurr.Opin.Environ.Sustain.3052–58 Environ.Health1595–111 LwasaS,MugaggaF,WahabB,SimonD,ConnorsJandGriffithC ScottAAetal2017Temperatureandheatininformalsettlements 2014Urbanandperi-urbanagricultureandforestry: inNairobiPLoSOne121–17 transcendingpovertyalleviationtoclimatechange SinghR,ArrighiJ,JjembaE,StrachanK,SpiresMand mitigationandadaptationUrbanClim.792–106 KadihasanogluA2019HeatwaveguideforcitiesTechnical MalchaireJ,KampmannB,HavenithG,MehnertPand Report (Geneva:RedCrossRedCrescentClimateCentre) GebhardtH2000Criteriaforestimatingacceptable (availableat:www.climatecentre.org/downloads/files/IFRC exposuretimesinhotworkingenvironments:areviewInt. Geneva/RCCC%20Heatwave%20Guide%202019%20A4% Arch.Occup.Environ.Health73215–20 20RR%20ONLINE%20copy.pdf) 11 Environ. Res. Lett.17(2022)024004 JVandeWalleetal SouchCandSouchC1993Theeffectoftreesonsummertime VandeWalleJ2021Canlocalfieldworkhelptorepresent belowcanopyurbanclimates:acasestudyBloomington, intra-urbanvariabilityofcanopyparametersrelevantfor IndianaJ.Arboric.19303–12 tropicalAfricanclimatestudies?Theor.Appl.Climatol. StewartID2011RedefiningtheurbanheatislandPhDThesis 146457–74 UniversityofBritishColumbiap352(availableat: VandeWalleJ,ThieryW,BrousseO,SouverijnsN,DemuzereM https://circle.ubc.ca/handle/2429/38069) andvanLipzigNP2020Aconvection-permittingmodelfor StewartIDandOkeTR2012Localclimatezonesforurban theLakeVictoriaBasin:evaluationandinsightintothe temperaturestudiesBull.Am.Meteorol.Soc. mesoscaleversussynopticatmosphericdynamicsClim.Dyn. 931879–900 541779–99 StewartIandKennedyC2015Estimatinganthropogenicheat VanLeeuwenJM,SekeramayiT,MartellC,FeinbergMand releasefrommegacitiesICUC9–9thInt.Conf.onUrban Bowersox-DalyS2017AbaselineanalysisoftheKatanga ClimateHeldJointlywiththe12thSymp.ontheUrban slums:informingurbanpublicpolicyinKampala,Uganda Environment pp20–24(availableat:www.meteo. Afr.Popul.Stud.312 fr/icuc9/LongAbstracts/gd5-1-2961311_a.pdf) VenterZS,BrousseO,EsauIandMeierF2020Hyperlocal TahaH1997Urbanclimatesandheatislands:albedo, mappingofurbanairtemperatureusingremotesensingand evapotranspirationandanthropogenicheatEnergyBuild. crowdsourcedweatherdataRemoteSens.Environ. 2599–103 242111791 ThieryW,DavinEL,PanitzH-J,DemuzereM,LhermitteSand VermeirenK,VanRompaeyA,LoopmansM,SerwajjaEand vanLipzigN2015TheimpactoftheAfricanGreatLakeson MukwayaP2012UrbangrowthofKampala,Uganda: theregionalclimateJ.Clim.284061–85 patternanalysisandscenariodevelopmentLandscapeUrban ThieryW,DavinEL,SeneviratneSI,BedkaK,LhermitteSand Plan.106199–206 VanLipzigNP2016Hazardousthunderstorm VermeirenK,VanmaerckeM,BeckersJandVanRompaeyA2016 intensificationoverLakeVictoriaNat.Commun. ASSURE:amodelforthesimulationofurbanexpansion 71–7 andintra-urbansocialsegregationInt.J.Geogr.Inf.Sci. ThieryW,GudmundssonL,BedkaK,SemazziFHM, 302377–400 LhermitteS,WillemsP,vanLipzigNPMand WilbyRLetal2021Monitoringandmoderatingextremeindoor SeneviratneSI2017Earlywarningsofhazardous temperaturesinlow-incomeurbancommunities thunderstormsoverLakeVictoriaEnviron.Res.Lett. monitoringandmoderatingextremeindoortemperatures 12074012 inlow-incomeurbancommunitiesEnviron.Res.Lett TwinomuhangiR,SseviiriH,MulindeC,MukwayaPI, 16024033 NimusiimaAandKatoAM2021Perceptionsand WMO1986Guidelinesontheselectionofreference vulnerabilitytoclimatechangeamongtheurbanpoorin climatologicalstationsfromtheexistingclilmatological KampalaCity,UgandaReg.Environ.Change211–13 stationnetwork(availableat:https://library.wmo.int/ UgandaBureauofStatistics2014UgandaNationalPopulationand doc_num.php?explnum_id=9310) HousingCensus(Kampala:UgandaBureauofStatistics WoodhamsBJ,BarrettPA,MarshamJH,BirchCE,BainC, (UBOS))(availableat:https://unstats.un.org/unsd/demogra FletcherJ,HartleyAJandWebsterS2021Firstaircraft phic/sources/census/wphc/Uganda/UGA-2016-05-23.pdf) observationsofthelakeVictorialake–landbreeze UnitedNations2018WorldUrbanizationProspects2018: circulationfromtheHyVicpilotflightcampaign34thConf. Highlights(ST/ESA/SER.A/421)(availableat: onHurricanesandTropicalMeteorology(AMS)(availableat: https://population.un.org/wup/Publications/ https://ams.confex.com/ams/34HURR/meetingapp.cgi/ Files/WUP2018-Highlights.pdf) Paper/386612) UnitedNations2019WorldPopulationProspects2019:Highlights XuZandTongS2017Decomposetheassociationbetween (DepartmentofEconomicandSocialAffairs,Population heatwaveandmortality:whichtypeofheatwaveismore Division)(availableat:https://population.un.org/wpp/ detrimental?Environ.Res.156770–4 Publications/Files/WPP2019_Highlights.pdf) ZscheischlerJetal2018Futureclimateriskfromcompound vandeGiesenN,HutRandSelkerJ2014TheTrans-African eventsNat.Clim.Change8469–77 Hydro-MeteorologicalObservatory(TAHMO)WIREs ZscheischlerJetal2020Atypologyofcompoundweatherand Water 1341–8 climateeventsNat.Rev.EarthEnviron.1333–47

Journal

Environmental Research Letters – IOP Publishing

Published: Feb 1, 2022

Keywords: heat stress; Humidex index; normalized difference vegetation index; local climate zones

There are no references for this article.