Earth observation for drought risk financing in pastoral systems of sub-Saharan Africa
Francesco Fava
1and Anton Vrieling
2Asclimate-relatedcrisesincreaseglobally,climaterisk financingisbecominganintegralpartoffinancialprotection andresiliencebuildingstrategiesofAfricancountries.Drought- inducedcrisesresultindevastatinghumanimpactsandhigh costsforvulnerablecountries,threateninglonger-term investmentsanddevelopmentefforts.Whileearthobservation (EO)hasbeenwidelyusedfordroughtearlywarning,new opportunitiesemergefromintegratingEOdataandmethods intoindex-baseddroughtriskfinancing(IBDRF)instruments.
Suchinstrumentsaimatsupportinganeffectiveandtimely responseduringdroughtshocksandimprovingtheresilience ofsmall-holderfarmersandlivestockkeepers.Thisreview documentsthecurrentstatus,anddiscussesfutureprospects andpotentialchallengesforEOutilizationinIBDRF
applicationsinsub-SaharanAfrica.Wefocusonpastoral systems,whicharehotspotsintermsofvulnerabilitytoclimate andenvironmentalchange,foodinsecurity,poverty,and conflicts.Inthesesystems,EO-basedIBDRFinterventionsare rapidlyscalingupaspartofnationalandinternationalrisk managementstrategies.
Addresses
1InternationalLivestockResearchInstitute(ILRI),P.O.Box30709, Nairobi00100,Kenya
2UniversityofTwente,FacultyofGeo-informationScienceandEarth Observation,P.O.Box217,7500AEEnschede,TheNetherlands
Correspondingauthor:Fava,Francesco(f.fava@cgiar.org)
CurrentOpinioninEnvironmentalSustainability2021,48:44–52 ThisreviewcomesfromathemedissueonThedrylandsocial- ecologicalsystemsinchangingenvironments
EditedbyBojieFu,MarkStafford-SmithandChaoFu ForacompleteoverviewseetheIssueandtheEditorial Availableonline18thOctober2020
Received:01June2020;Accepted:14September2020 https://doi.org/10.1016/j.cosust.2020.09.006
1877-3435/2020TheAuthor(s).PublishedbyElsevierB.V.Thisisan openaccessarticleundertheCCBYlicense(http://creativecommons.
org/licenses/by/4.0/).
Introduction
Crisis risk financing refers to mechanisms that aim at reducingadversesocio-economicorecologicalimpactsof potentialcrises[1].Thiscanincludepaying toprevent
and reduce therisk, or to prepare for and respond to a shock. Climate risk financing (CRF) targets climate- related shocks (e.g. drought, floods and heat waves) andisbecominganintegralpartofclimateriskmanage- mentframeworksaskeycomponentsoffinancialprotec- tionstrategicplanningforlowandmiddleincomecoun- tries [2]. Multiple CRF approaches exist, including market-basedinstruments(e.g.insuranceschemes,cata- strophic bonds and swaps), contingent financing (e.g.
credit), or budgetary tools (i.e. dedicated reservefunds or contingency budgets). These approaches are all designedtoincreasefinancialresiliencetoclimate-related crises,linkingtheresponseactionstopre-definedmech- anisms for timely release of financial resources. In this way,theyaimatensuringrapidandcost-effectiveprepa- ration,assistance,recoveryor reconstructionefforts.
Droughts,definedasperiodswithwaterdeficitrelativeto normalconditions,areoneofthemostdisruptingnatural disasters, each year affecting millions of people world- wide with devastating impacts [3,4]. Severe droughts cause massive disruptions to national economies and dramaticimpactson thelivelihoodand foodsecurity of small-holder farmers and livestock keepers. Standard responsestodroughtinAfricancountries,suchashuman- itariansupportin theformofcash or foodtransfers,are important instruments to support drought-affected vul- nerable populations. However, these responses have proventobeoftentooslow,cost-ineffective,andtofoster dependencyrather than resilience, especiallywhennot integratedintoholisticrisk managementstrategies[2].
Among the different CRF instruments, index-based approacheshavegainedconsiderabletractionoverthelast two decades, particularly for targeting drought shock impactsonAfricansmall-holderfarmingsystems.Index- baseddroughtriskfinancing(IBDRF)usestriggermecha- nismsthatrelyonatransparentandobjectivelymeasured indicatorofdrought(i.e.theindex).Theunderlyingindex mustbehighlycorrelatedwithdrought-relatedeconomic lossestobeusefulintracking,and,therefore,transferring, therisk.Payoutsaremadewhentheindexvaluesfallbelow a pre-defined threshold, normally derived from historic indexrealizations.Whencomparedtoex-postlossverifica- tion(e.g.traditionalinsurance),IBDRFlimitsinformation asymmetry issues, such as adverse selection and moral hazards,3reduces transaction and verification costs, and
3Thisterminologyoriginatesfrominsuranceliterature:adverseselectionoccurswhenpotentialpolicyholdersmakedecisionsbasedoninformation abouttheirriskexposurethatisnotavailabletotheinsuranceprovider.Moralhazardoccurswhenthepolicyholdersengageinactivitiesthatincrease theirexposuretorisk,leavingtheinsurerexposedtohigherriskthanhadbeenassessedforpremiumratedetermination.
enablesmoreeffectiveandtimelydistributionofpayouts.
However,theseadvantagescomeatthecostoftheinher- entlyimperfectcorrelationbetweenindexandactualloss, alsoknownasbasisrisk[5].
Pastoral production systems dominate the African dry- lands,coverabout43%ofAfrica’slandmass,andarethe main livelihood for about 268 million people in these areas[6].Droughtis adistinctivefeatureand over mil- leniathepastorallivelihooddevelopedto dealwith this [7]. However, particularly in SSA, the combination of increasingrainfallvariability[8]andotherpressures(e.g.
changes in land use and tenure, population growth, rangeland degradation [9]) is weakening the resilience of pastoral communities and the effectiveness of tradi- tional drought coping mechanisms (e.g. mobility, re- stocking).Severedroughtscanleadtocatastrophicherd lossesinpastoralregions,andassuchcausefoodinsecu- rityandpovertytrapdynamics,anddramaticallyreduce national GDPsof SSA countries[5].Consequently, the development and implementationof drought riskman- agementstrategiesforpastoralregions,includingIBDRF initiatives, isakey component of thepolicyagendasof developmentinstitutions andnationalgovernments.
Thescarcityandpoorqualityofgrounddataandnational statistics in SSA pastoralregions[10] hasmadein most casesEarthobservation(EO)bysatellitestheonlyviable optionfor designing IBDRFinstruments for large-scale implementation. Thegrowingavailability andincreased qualityoflong-termEOdatasetsofrainfallproductsand vegetation indices [11] have been instrumental in designing indicesthat are thought to reliably represent drought risks. Furthermore, the close interlinkage betweenrangelandconditionandimpactsonproductive assets (i.e. livestock)and livelihoods has facilitated the designoffinancialtriggeringmechanismsthatlinkimpact to payouts.Therefore,datasetsobtainedthroughmulti- temporalsatelliteimageryarecurrentlyakeycomponent of IBDRFinitiativesin pastoralregions.
However,whilstthescientificliteratureondroughtmon- itoringisvast,onlyafewstudiesreviewedthecontribu- tion ofEO toinsurance[12,13],andcurrentlynoover- view exists of the IBDRF initiatives, challenges, and prospectsin sub-SaharanAfrica.
Evolution ofIBDRF inpastoral Africa
During the last decade IBDRF initiatives in African pastoraldrylandshavegainedsignificantmomentum,also thankstolargeinternationalinitiativessuchastheInsuR- esilience Investment Fund [14] and the Global Index- InsuranceFacility(GIIF).WhileforseveralyearsIBDRF schemes have largely remained at the pilot level, con- strainedbythelimiteddemandforretailmicro-insurance products[15,16],theseschemesarenowgraduallycover- ing larger areas, given their growing integration in
country-widesocialprotectionanddroughtriskmanage- ment programs. This caused a significant increase in volumeoffinancialtransactionsandagrowingparticipa- tionofinternationalre-insurancecompanies,whichfacil- itatesrisktransfertowardinternationalfinancialmarkets, and promotes larger investments by governments and internationalorganizations.
Table 1 summarizes operational IBDRF initiatives in SSA.. These initiatives, initially launched as retail micro-level index-insurance schemes (IBLI), more recently have expanded their scope and modality of implementation, including fully subsidized insurance programstargetingvulnerablepastoralists(KLIP,SIIPE), sovereign-levelinsuranceschemes(ARC),andscalability mechanisms of shock responsive safety nets programs (HSNP, NUSAF). At the same time, IBRDF product design evolved from indices designed to assess an observedloss(i.e.theIBLIlivestockmortalityscheme), thustriggeringwhendroughtisalreadyimpactingpasto- ralistassets,toindicesdesignedtoidentifydeteriorating forage condition, thus triggering at the earlier drought stages,withtheaimofsupportingpastoraliststo protect their assets (i.e. livestock or livelihood) or countries implementingearlyresponseactions.
While fully operational initiatives in pastoral areas are concentratedlargelyinEastAfricaand,toalesserextent, in the Sahel, several insurance providers and organiza- tions arelaunching similar IBDRFsolutions on aretail basisacrossSSA, includinginNiger, ZambiaandSouth Africa.Inaddition,feasibilitystudiesarebeingconducted in Somalia,Senegal,Burkina Faso,andMali. Thusles- sons gatheredfrom early implementation efforts are of keyrelevancefor sustainablescaling ofIBDRFin SSA.
EO-based index designinoperationalIBDRF schemes
Similartomostdroughtearlywarningsystems[e.g.Refs.
17,18,19],existingIBDRFschemesforpastoralsystems predominantlyusevegetationindicesorrainfallestimates (RFE) as input (Table 1). These EO products should meet the fundamental operational requirements for IBDRF,whichinclude 1)fulltransparencyandaccessi- bilityofthesourcedata,2)availabilityofhistoricalrecords forfinancialriskmodellingandpricingofideally20years ormore,3)nearreal-timeavailability,and4)anexpected remaining lifetime of at least a few years. Satellite- derivedRFEcanprovideadirectindicationofmeteoro- logicaldrought,whichlargelydetermineswateravailabil- ityin rangelands.While availabilityof drinkingwateris important for livestock health, monitoringthis for large areasiscumbersome,giventhatmuchwaterforlivestock comesfromwellsorsmallwaterbodiesthatcannoteasily bemonitoredwithEO data. RFEsare insteadtypically usedinIBDRFtoestimatetheavailablewaterforvege- tation, for exampleusing simple water-balance models.
Thewater requirementsatisfactionindex (WRSI)used byARCissuchamodel;itestimatesevapotranspiration demandsofvegetationandcomparesthiswithasimple
‘bucket’ model of the soil that gets filled with rainfall [20,21].Twomaindrawbacksexistwhenusingrainfall- basedindicesforlarge-scalepastoralIBDRF:1)despite thegrowingavailabilityofgriddedrainfallproducts,the quality,accessibility,anddensityofrainfallstationdata in pastoral SSA is generally low [22,23], resulting in unknown orpoor accuraciesin these areas [24]; 2)the link between rainfall and the vegetation’s water avail- abilityiscomplex,anddependsonvegetationcharacter- istics,soil,andrainfalldistribution[25,26],whichcannot becharacterizedsufficiently with10-daily rainfallsums [27].
For extensive pastoral systems, forage availability is a keydeterminantoflivestock health,asalternative feed resources are largely unavailable or inaccessible. To overcome drawbacks of RFEs, optical sensors onboard satellitescanbeusedtomeasurethevegetation’sreflec- tance. Dense healthy vegetation reflects much in near infrared(NIR),andlittleinredwavelengths,and spec- tral vegetation indices, like the normalized difference
vegetationindex(NDVI),usethistomonitorvegetation condition. Fordroughtmonitoring, NDVI imagesfrom coarse-resolutionsatellites(250mandup)aregenerally used,because1)longtimeseriesexisttodescribelong- termvariability inforageconditions,2)their dailydata collections allow for more cloud-free observations to describe vegetation changes throughout the season, and3)documentedevidenceexistsforastrongrelation- ship between rangeland biomass and NDVI [28,29].
Requiredpre-processingstepstoeffectivelyuseNDVI for anomaly analysis include temporal compositing [transforming daily to, e.g. 10-day data, keeping the best-qualityobservationforeachpixel;30],andsmooth- ingtofurtherreduceatmosphericeffects [31].
TotransformNDVIoralternativedroughtindicesintoa usefulindexforpastoralIBDRFschemes,threestepsare required:
1) Spatial aggregation; geographic units are normally larger than grid cells, both for operational reasons and to reflect that herds move. Aggregation within units generally incorporates amask of where range- landsoccur.
Table1
SummaryofoperationalIBDRFprogramstargetingpastoraldrylands,orderedchronologicallywithinsuranceprogramslistedfirstand non-insuranceprogramsatthebottom
Programa Years Scope Satelliteb
sensor/indicator
Indexc Area Households(nr)
IBLI 2010 2015 Retailmicro- insurancefor droughtrelated livestockmortality
AVHRRNDVI Livestockmortality Marsabit(Kenya), Borana(Ethiopia)
5983
IBLI 2015–present Retailmicro- insurancefor assetprotection
MODISNDVI z-scoreseasonalNDVI NorthernKenya andBorana (Ethiopia)
7000
KLIP 2015–present Subsidized insurancefor assetprotection
MODISNDVI z-scoreseasonalNDVI Northernand EasternKenya
18000
SIIPE 2017-present Subsidized insurancefor assetprotection
MODISNDVI z-scoreseasonalNDVI Somaliregion (Ethiopia)
5000
ARC 2017-present Sovereignlevel insurance
RFE–multiple datasets
WRSI(RFEbased) EastAfricaand Sahel
N/ad ARC 2019-present Sovereignlevel
insurance
MODISNDVI z-scoreNDVIorVCI EastAfricaand Sahel
N/ad
HSNP 2015-present Socialprotection scalability mechamism
MODISNDVI VCIrunningaverage NorthernKenya >100000
NUSAF 2017-present Socialprotection scalability mechanism
MODISNDVI NDVIpercentanomaly Karamoja (Uganda)
25000
aIBLI=Index-basedLivestockInsurance,KLIP=KenyanLivestockInsuranceProgram,SIIPE=SatelliteIndex-InsuranceforPastoralistsinEthiopia, ARC=AfricanRiskCapacity,HSNP=HungerSafetyNetProgram,NUSAF=NorthernUgandaSocialActionFund.
bAVHRR=AdvancedVeryHighResolutionRadiometer.MODIS=ModerateResolutionImagingSpectroradiometer,NDVI=NormalizedDifference VegetationIndex,RFE=RainfallEstimates(usingsatellitedata).
cZ-scoreissometimesreferredtoasstandardscore,WRSI=WaterRequirementSatisfactionIndex,VCI=VegetationConditionIndex.
dAsARCisasovereign-levelinsurancescheme,therecipientofthepayoutisthecountryandthenumberofhouseholdscoveredwilldependonthe country’scontingencyplans.ARCisofferinginsurancecoverforrangelandsintheentireSahelandHornofAfricaandplanstoofferitalsoinCentral andSouthernAfrica.TheNDVIproducthasbeenlaunchedinChad,Niger,Mali,Mauritania,andKenya.
2) Temporal aggregation; most schemes aim to assess seasonal forage scarcity, requiring expert or EO- derived [32] knowledge on rainfall/vegetation seasonality.
3) Normalization to compare the current index value againsthistoricindexrealizationsinpastyears.Multiple options exist, suchas forexamplez-scoring (subtract meananddividebystandarddeviation),linearscaling between minimum and maximum historic values [i.e. the vegetation condition index, or VCI; 33], or percentilecalculation.
Theterm‘indexdesign’inIBDRFrefersto:a)thechoice ofinputdata,b)theprecisemethodsforperformingthe above-mentionedthreesteps,c)thesequencingofthese steps, and d) the testing of the resulting index against drought-relatedlossestimates.Eventually,thechoicefor a design will at least partly depend on the IBDRF instrument’spurpose,satisfactionofstakeholdersonhis- toric and current index readings, and (scientific) back- groundof the‘designers’.Giventhatdroughtsnormally affectlargeregionsandaresimultaneouslycharacterized byreducedprecipitationandvegetationgrowth,itiswell possible that alternative designs provide similar index outcomesduringmaindroughtevents.
Opportunities and challengesforEO contribution to IBDRF
RecenttargetedresearcheffortsforEOsupporttoinno- vativedesignofIBRDFsolutions,madeintheframework of the programs listed in Table 1, helped to remove criticalbarriers for scalingIBDRFinitiatives in pastoral drylands. Here, we discuss four emerging trends in IBDRFwherethedemandforinnovativesolutionsfrom EOresearchisstrong.Whilewearguethatthequalityof IBDRF product design differsand thatEO-aided solu- tions canplayanimportant role to improve operational programs, we also emphasize that each solution iscon- text-specificandneedstoconsiderthetrade-offbetween a)timingand accuracyof theassessment, b)cost-effec- tivenessoftheresponseaction,andc)thecomplexityof the interaction between drought shocks and the socio- ecological pastoral systems (Figure 1) with the goal of mitigatingimpactsandspeeding-uprecoveryintheshort term,whilebuildinglong-termsocialandenvironmental resilience.
EmergingEOdataproductsforIBDRF
EO data products used for IBDRF index design in pastoral areas have evolved over time in response to stakeholder feedbacksand with emergingtechnologies.
For example,theARC productforextensiverangelands hasrecentlytransitionedfromWRSItoNDVI(Table1), meeting the demand of key stakeholders. However, besidesvegetationindex(e.g.NDVI)productsthatpro- videadirectmeasureofforagestatus,droughtcharacter- istics can be obtained, for example, from EO-derived
precipitation[35],soilmoisture[36,37],orevapotranspi- ration[38]products[forreviewsonEOdroughtmonitor- ing options see also Refs. 11,39]. A variety of these productshavealreadybeen proposedandinsomecases integratedintodroughtindex-insurancepilotprojectsfor crops inAfrica,for example,throughdataserviceprovi- sion by commercial EO companies [e.g. Refs. 40–43].
Theseeffortsprovidepromisingalternativedroughtindi- cesforinnovativeIBDRFdesignbyaddressingmultiple phases of drought progression and thus merit further analysis.
IBDRFdesigncouldalsotakeadvantageofthecontinu- ous streamof free10 30m resolutionEO datathatare providedbysatellitessuchasSentinel-1,Sentinel-2,and Landsat-8. Given their high observation frequency, timelyfine-scaleestimatesofseasonalityandforagecon- ditionscanbeprovided[44],evenifcloudcoverremainsa concernfor short vegetationseasons [45].Arguably pas- toral IBDRF may not require fine-scale data because droughtsgenerallyaffectlargeareas, butwheredrought impactsdifferduetogreaterlandscapevariability(e.g.in agro-pastoral systems) they could prove beneficial. As historical archivesof 10 30mresolution dataarebuild- ing, and tools to analyze resulting large data amounts become commonplace [46], finer-scale droughtanalysis will likelyfinditswayintoIBDRF.
Remotesensingadvancesforimprovedproductdesign Thespatialcomponentofbasisrisk(i.e.impactsarenot equallydistributedwithinageographicinsuranceunit)is arecurrentissueforIBDRFproductsinpastoralareas,as administrativeunitscannotfullyreflecttheagro-ecologi- cal variability and herd mobility patterns. This type of basisriskhasbeenreportedfor someinsuranceunitsin Kenya and Ethiopia, especially at the fringe between agro-pastoral and extensive pastoral systems. To deal with this issue, administrative units could be replaced by more meaningful ecological units, for example, defined based on similar temporal behavior of NDVI [47]. In addition, high resolution EO data can help to improve rangeland mapping and characterization [48], allowing to better spatially focus coarser-resolution droughtindicesontheareaswithininsuranceunits that mattermostforforageproduction.
Payouttimingandtemporalaggregationisanothercritical componentofproductdesign.Triggeringearlyactionfor expectedadverseeventscanmakedisasterresponsemore cost-effective [49,61]. Because drought is aslow-onset shock,indicatorscanbedesignedthatdetectearlystages of thedrought progression[11].EO has supportedthe anticipationoftheresponsebyeffectivelycharacterizing between-yearvariabilityofforageconditionsearlierinthe season through shortening temporal aggregation win- dows, with the aim of designing assetprotection insur- ance products[49,62,63].Thishasbeenafundamental
step for the IBLI and KLIP programs to increase the uptakeandgainpoliticalsupporttowardgeographicscal- ing.However,indicesthatallowforearlytriggeringmust alsobebackedbyefficient payoutdeliverymechanisms toavoidnegative impactsonprogram sustainability.
Morerecently,forecastingapproachesbasedontimeseries analysisandmachinelearningtechniqueshavebeendevel- opedtopredicttime-integrated(seasonal)NDVIanomalies fromlagged NDVI,rainfall,andclimateindices[64–68].
Whileresearch ondroughtforecasting modelsisgaining momentum, particularly for early warning systems (e.g.
HSNP, Table 1), the implementation of forecasts into IBDRFinstruments is challenginggiventheforecasts’ large uncertaintyatlocalscalesandlongertimelags[69].Using forecastsasatriggeringindexcouldthusincreasebasisrisk andleadtooperationalimplementationchallenges(e.g.in caseoffalsealarmandunneededresponse)[61,70].More- over,decidingonthresholdsforresponsetriggeringcould becomplexbecausethisrelatesnotonlytothelevelofthe impact, but also to itsprobability. A differentchallenge emerges when forecastsare not directly used but could nonethelessaffectIBDRFinstrumentssuchasinsurance.
Forexample,sinceinsurancepremiumsarebasedonhis- toricalrealizationsoftheindex,reasonablyaccuratefore- castsmadebeforeinsurancesales[possiblythroughindige- nous indicators: 71] could lead to adverse selection if prospectivepurchasers have informationabout expected payoutsthattheproduct’spricingdoesnotaccountfor[e.g.
Ref.15].Overall,whileanticipatoryriskfinancingbasedon forecastsisapromising innovationin IBDRF, itsopera- tionalimplementationwouldrequireacarefulassessment ofassociatedrisksandeffectiveharmonizationwithother IBDRFinstruments.
QualityassuranceofIBDRF
Basis riskremains acriticalconcern for thequality and sustainabilityofIBDRFschemes.Foroperationalinitia- tivesafewrecentstudiescomparedresultsfrommultiple index designs [41,47,49,50]. Notwithstanding, these studieshighlightedthatevaluationoftheresultingindi- cesremainscomplexgiventhescarcityofand/ordifficulty tocollectgood-qualityin-situdataondroughtoutcomes.
While EO product development and improved index design havepotentialto reducebasis risk,ultimately it remainsan empirical questionwhether this potentialis
Figure1
DROUGHT PROGRESSION IN PASTOROL SOCIO- ECOLOGICAL SYSTEMS
TIMING &
ACCURACY
COST-
EFFECTIVENESS COST OF RESPONSE
TYPE OF RESPONSE
ACCURACY OF ASSESMENT / BASIS RISK TRIGGER TIMING
Current Opinion in Environmental Sustainability
GroundandsatelliteEOapplicationscancontributetomonitordroughtprogressionanditsecologicalandsocio-economicimpactstosupport context-specificdesignofIBDRFinstruments.TheupperpartoftheFigureillustratestheprogressionofdroughtimpactsfollowingMishraand Singh[34].Whiledroughtevolvesfrommeteorologicaltosocio-economic,IBRDFdesignneedstoconsidertrade-offstomaximizethesocial- ecologicalbenefits,whileaccountingforeconomiccoststoascertainlong-termsustainabilityoftheinstruments.Forexample,anEO-basedindex designedwithanearlytriggerisexpectedtobemoreaccurateindetectingmeteorologicaloragriculturaldrought,butmightbelesseffectivein assessingsocio-economicimpacts.Thissourceofbasisriskmay(ormaynot)becounterbalancedbythesavingsineconomiccostsassociated withassetprotectionandearlyresponse.Theseassumptionsarecontext-specificandproductqualityassessmentisthusfundamentaltoevaluate trade-offs.
realized.Thisquestiongoesbeyondthetraditionalaccu- racy assessment approachesin the EO domain (e.g. an evaluationifsoilmoistureisaccuratelyrepresented ina soil moisture product),as it should encompass also the socio-economicvalueoftheproposedsolution[49],and be formalized through quality metrics, minimum stan- dards and robust assessment frameworks [1,51]. The overarchinggoalshouldbetodesignrigorousquantitative metricsandapproachestoassessandcomparetheutility of IBDRFinterventions.
A main challenge is collecting and analyzing relevant reference data on drought outcomes, given the limited formal data sets [52]. Potential data sources include multi-year forage biomass measurements [53], drought recallexercises[52],andrepetitivehouseholdsurveyson droughtoutcomessuchaslivestockmortality[49]orchild nutrition[54].Giventhecost-intensiveandlabor-intensive nature of collecting such data, ground-based EO approachescanprovideausefuladdition.Agoodexample istherepetitiveobservationofthesamevegetation,either bypermanentcameras[55,56],orthroughcrowdsourcing platforms[57].Forcrops,thishasalreadyledtotheideaand implementation of ‘picture-based-insurance’ whereby farmersrepeatedlycollectpicturesoftheirfieldsforverifi- cation of insuranceclaims[58]. Thisidea could be extended to rangeland and livestock conditions, possibly taking advantage of computer-visionbased automationof body scoringtechniques[59,60].Increasedeffortsinreference datacollectionareurgentlyneededtoanswertheempirical questionandprovidequalityinsuranceproducts.
Socio-economicandenvironmentalimpactevaluation With the geographic expansion and growing number of households covered, the need for impact evaluation of IBDRFinitiativeshasincreased.Traditionallyimpacteval- uationofdroughtcrisesonsmall-holderfarmersandlive- stock keepers largely focused on socio-economic factors through mixed qualitative-quantitative household-level surveys [72].However,the highcosts andcomplexityof such surveys haveprevented asystematic integrationof impactassessmentstudies in IBDRF initiatives, so that only fewrobustassessmentsareavailableinpastoralregions[e.g.
Ref.73].Inaddition,environmentalimpacts have been poorlyconsidered so far, while scalingIBDRFinterventions to a large number of households might have relevant ecological impacts. For example, reduced herd losses may result in increased grazing pressure on rangelands [74,75],althoughempiricalstudiessuggestonthecontrary thatinsuredpastoralistskeepsmallerherdsbecausethey reducetheiruseoflivestockasprecautionarysavings[76].
EOapproachesforrangelandhealthmonitoringarewell established [e.g. Ref. 77]. However, only few studies, focused on land restoration, have integrated these approachesinto impactassessment of largeprogramsin drylands[78,79].Recentliteraturealsoshowedpotential
for assessing food security and household wealth [80,81,82], but pointed out that satellite EO capacity to directly monitor socio-economic indicators of house- hold wellbeing and rapid land use dynamics during drought events (e.g. land accessibility, land tenure changes, migration)is limited. Impact evaluation could beimprovedbycombiningsatelliteEOinformationwith in-situcollectedenvironmentalandsocial-economicdata [83], which are increasingly available via mobile and crowdsourcing platforms also in rural Africa [e.g. Refs.
84,85,86].Thiscanbesupportedbyadvancesinmachine learning algorithms, allowing to extract patterns and understandspatialconnectionsfromdiversedatasources [87]. Datadriven approachesshould be, however, used with caution and framed within a robust set of causal hypotheses, taking into account the cross-scaleinterac- tionsbetweenphysicalandsocio-economicfactors[81].
Conclusions
IBDRF initiatives are scaling up as part of the policy agendasofSSAcountriesanddevelopmentorganizations onresiliencebuilding,povertyreduction,andsustainable development.EOplaysakeyroleinsustainingthistrend in pastoral drylands, with the potential of significant societalandpolicyimpactintheregion.Innovationsfrom groundandsatelliteEOtechnologiescouldcontributeto design more accurate, cost-effective, and harmonized droughtriskfinancingprograms,aswellastoassesstheir effectiveness during shocks and their longer-term impacts. However,thiscan onlybeachievedif theEO communitydoesnotlimititsroletoprovidingtechnolog- icalsolutionsandservices,butratherbecomesanintegral partoftheinterdisciplinaryframeworktoaddressdrought risk management in complex socio-ecological systems, understandingsynergiesandtrade-offsbetweenresearch, operationalimplementation,and policyformulation.
Conflictofinterest statement Nothingdeclared.
Acknowledgements
TheCGIARResearchprogramonLivestockandcontributorstothe CGIARTrustFundareacknowledgedforsupportingthisresearch.Anton VrielingwasfundedbytheDutchResearchCouncil(NWO),Spacefor GlobalDevelopment(WOTRO)programme,aspartoftheCGIAR- Netherlandspartnership.WethankNathanielJensen(ILRI)andJames Hassell(SmithsonianGHP)fortheirvaluablecommentsandinputsonthe manuscript.
References and recommendedreading
Papersofparticularinterest,publishedwithintheperiodofreview, havebeenhighlightedas:
ofspecialinterest
1. PooleL,ClarkeD,SwithernS:TheFutureofCrisisFinancing:ACall toAction. London,UK:CentreforDisasterProtection;2020 https://www.disasterprotection.org/crisisfinance.
Thereportprovidesanoverviewofcrisisfinancingandareferencefor terminology, setting out a new vision for effective crisis financing packagesandinstruments.
2. TheWorldBank:DisasterRiskFinanceasaToolforDevelopment:
ASummaryofFindingsfromtheDisasterRiskFinanceImpact AnalyticsProject. WashingtonDC:TheWorldBank;2016https://
openknowledge.worldbank.org/handle/10986/24374.
3. Carra˜oH,NaumannG,BarbosaP:Mappingglobalpatternsof droughtrisk:anempiricalframeworkbasedonsub-national estimatesofhazard,exposureandvulnerability.GlobalEnviron Change2016,39:108-124.
4. MezaI,SiebertS,Do¨llP,KuscheJ,HerbertC,RezaeiEE,NouriH, GerdenerH,PopatE,FrischenJetal.:Global-scaledroughtrisk assessmentforagriculturalsystems.NatHazardsEarthSystSci 2020,20:695-712.
5. BarrettCB,BarnettBJ,CarterMR,ChantaratS,HansenJW, MudeAG,OsgoodD,SkeesJR,TurveyCG,WardMN:Poverty TrapsandClimateRisk:LimitationsandOpportunitiesofIndex- BasedRiskFinancingIRITechnicalReportNo.07-02.2007.
6. FAO:PastoralisminAfrica’sDrylands;ReducingRisks,Addressing VulnerabilityandEnhancingResilience. Rome,Italy:Foodand AgricultureOrganizationoftheUnitedNations;2018.
7. ButtB,ShortridgeA,WinklerPrinsAMGA:Pastoralherd management,droughtcopingstrategies,andcattlemobilityin SouthernKenya.AnnAssocAmGeogr2009,99:309-334.
8. SloatLL,GerberJS,SambergLH,SmithWK,HerreroM, FerreiraLG,GoddeCM,WestPC:Increasingimportanceof precipitationvariabilityongloballivestockgrazinglands.Nat ClimateChange2018,8:214-218.
9. HallerT,VanDijkH,BolligM,GreinerC,SchareikaN,GabbertC:
Conflicts,securityandmarginalisation:institutionalchangeof thepastoralcommonsina’glocal’world.OIERevue ScientifiqueetTechnique2016,35:405-416.
10. ZezzaA,FederighiG,KalilouAA,HiernauxP:Milkingthedata:
measuringmilkoff-takeinextensivelivestocksystems.
ExperimentalevidencefromNiger.FoodPolicy2016,59:174- 186.
11. WestH,QuinnN,HorswellM:Remotesensingfordrought monitoring&impactassessment:progress,pastchallenges andfutureopportunities.RemoteSensEnviron2019,232 111291.
Acomprehensivereviewsummarizingthelatestdevelopmentsofremote sensingtechnologiesfordroughtmonitoringandassessment,withaclear vision toward the future roleof new missions, sensor technologies, algorithmsandanalytics.
12. deLeeuwJ,VrielingA,SheeA,AtzbergerC,HadguK,BiradarC, KeahH,TurveyC:Thepotentialanduptakeofremotesensing ininsurance:areview.RemoteSens2014,6:10888-10912.
13.
VroegeW,DalhausT,FingerR:Indexinsurancesforgrasslands –areviewforEuropeandNorth-America.AgricSyst2019, 168:101-111.
Thisreviewsprovidesanoverviewofemergingtrendsinindex-insurance forgrasslandsinEuropeandNorthAmerica,settingthestageforfuture opportunities.
14. InsuResilience:JointStatementontheInsuResilienceGlobal Partnership.14November2017.Bonnhttps://www.insuresilience.
org/wp-content/uploads/2017/11/
20171205-Joint-Statement-InsuResilience-Global-Partnership.
pdf:Accessed:29May2020.
15. JensenND,MudeAG,BarrettCB:Howbasisriskand spatiotemporaladverseselectioninfluencedemandforindex insurance:evidencefromnorthernKenya.FoodPolicy2018, 74:172-198.
16. TakahashiK,IkegamiM,SheahanM,BarrettCB:Experimental evidenceonthedriversofindex-basedlivestockinsurance demandinsouthernEthiopia.WorldDev2016,78:324-340.
17.
RemboldF,MeroniM,UrbanoF,CsakG,KerdilesH,Perez- HoyosA,LemoineG,LeoO,NegreT:ASAP:anewglobalearly warningsystemtodetectanomalyhotspotsofagricultural productionforfoodsecurityanalysis.AgricSyst2019,168:247- 257.
Thisstudydescribesanewglobalearlywarningsystem,whichprovides timelywarningsthroughauser-friendlypubliconlinewebGIS.State-of- theartEOmethods areusedtoderiveagriculturalproductiondeficit
warning categories, which are integrated by comprehensive expert analyses.
18. FritzS,SeeL,BayasJCL,WaldnerF,JacquesD,Becker-ReshefI, WhitcraftA,BaruthB,BonifacioR,CrutchfieldJetal.:A comparisonofglobalagriculturalmonitoringsystemsand currentgaps.AgricSyst2019,168:258-272.
19. Becker-ReshefI,JusticeC,BarkerB,HumberM,RemboldF, BonifacioR,ZappacostaM,BuddeM,MagadzireT,ShitoteC etal.:Strengtheningagriculturaldecisionsincountriesatrisk offoodinsecurity:theGEOGLAMcropmonitorforearly warning.RemoteSensEnviron2020,237.
20. VerdinJ,KlaverR:Grid-cell-basedcropwateraccountingfor thefamineearlywarningsystem.HydrolProcess2002,16:1617- 1630.
21. TarnavskyE,ChavezE,BoogaardH:Agro-meteorologicalrisks tomaizeproductioninTanzania:sensitivityofanadapted WaterRequirementsSatisfactionIndex(WRSI)modelto rainfall.IntJApplEarthObservGeoinf2018,73:77-87.
22. NicholsonSE,FunkC,FinkAH:RainfallovertheAfrican continentfromthe19ththroughthe21stcentury.Global PlanetaryChange2018,165:114-127.
23. MaidmentRI,GrimesD,BlackE,TarnavskyE,YoungM, GreatrexH,AllanRP,SteinT,NkondeE,SenkundaSetal.:Anew, long-termdailysatellite-basedrainfalldatasetforoperational monitoringinAfrica.SciData2017,4170063.
24. NicholsonSE,KlotterD,ZhouL,HuaW:Validationofsatellite precipitationestimatesovertheCongoBasin.JHydrometeorol 2019,20:631-656.
25. WilcoxKR,ShiZ,GherardiLA,LemoineNP,KoernerSE, HooverDL,BorkE,ByrneKM,Cahill JJr,CollinsSLetal.:
Asymmetricresponsesofprimaryproductivitytoprecipitation extremes:asynthesisofgrasslandprecipitationmanipulation experiments.GlobalChangeBiol2017,23:4376-4385.
26. ZeppelMJB,WilksJV,LewisJD:Impactsofextreme precipitationandseasonalchangesinprecipitationonplants.
Biogeosciences2014,11:3083-3093.
27. KimaniMW,HoedjesJCB,SuZ:Anassessmentofsatellite- derivedrainfallproductsrelativetogroundobservationsover EastAfrica.RemoteSens2017,9:430.
28. SchucknechtA,MeroniM,KayitakireF,BoureimaA:Phenology- basedbiomassestimationtosupportrangelandmanagement insemi-aridenvironments.RemoteSens2017,9.
29. DingaanMNV,TsuboM:Improvedassessmentofpasture availabilityinsemi-aridgrasslandofSouthAfrica.Environ MonitAssess2019,191.
30. FensholtR,AnyambaA,StisenS,SandholtI,PakE,SmallJ:
Comparisonsofcompositingperiodlengthforvegetation indexdatafrompolar-orbitingandgeostationarysatellitesfor thecloud-proneregionofWestAfrica.PhotogrammEng RemoteSens2007,73:297-309.
31. CaiZ,Jo¨nssonP,JinH,EklundhL:Performanceofsmoothing methodsforreconstructingNDVItime-seriesandestimating vegetationphenologyfromMODISdata.RemoteSens2017,9.
32. ZengL,WardlowBD,XiangD,HuS,LiD:Areviewofvegetation phenologicalmetricsextractionusingtime-series,
multispectralsatellitedata.RemoteSensEnviron2020,237 111511.
33. KlischA,AtzbergerC:OperationaldroughtmonitoringinKenya usingMODISNDVItimeseries.RemoteSens2016,8:267.
34. MishraAK,SinghVP:Areviewofdroughtconcepts.JHydrol 2010,391:204-216.
35. SunQ,MiaoC,DuanQ,AshouriH,SorooshianS,HsuKL:A reviewofglobalprecipitationdatasets:datasources, estimation,andintercomparisons.RevGeophys2018,56:79- 107.
36. BabaeianE,SadeghiM,JonesSB,MontzkaC,VereeckenH, TullerM:Ground,proximal,andsatelliteremotesensingofsoil moisture.RevGeophys2019,57:530-616.
