DSCOVREPICVegetationEarthSystemDataRecord

ScienceDataProductGuideMaturitylevel:Provisional

Y.Knyazikhin,W.Song,B.Yang,T.ParkandR.B.Myneni

BostonUniversityBoston,April22,2018

1

TABLEOFCONTENTS

1.INTRODUCTION............................................................................................................................22.EXPERIMENTOVERVIEW...........................................................................................................4

3.VESDRSUN-SENSORGEOMETRY..............................................................................................5

4.PRODUCTTILING.........................................................................................................................65.LEVEL2VESDRPRODUCT..........................................................................................................6

5.1.VESDRproductfilename.........................................................................................................................65.2.HDFfilestructure........................................................................................................................................65.3.Rootattributes.............................................................................................................................................75.4.Datasets...........................................................................................................................................................85.5.Qualityassessmentdataset.....................................................................................................................9

6.ANCILLARYSCIENCEDATAPRODUCTS...............................................................................116.1.DSCOVREPIClandcovertype.............................................................................................................116.2.Distributionoflandcovertypes.........................................................................................................14

7.KNOWNISSUES..........................................................................................................................158.EXAMPLES...................................................................................................................................15

2

1.INTRODUCTION1.1. Purpose. This document describes Level 2 Vegetation Earth System Data Record(VESDR)derivedfromtheEarthPolychromaticImagingCamera(EPIC)onboardtheDeepSpace Climate Observatory (DSCOVR). It provides file structure for the geophysical andancillarysciencedataproducts.TheVESDRparametersaresummarizedinTable1.Table 1: Vegetation Parameter Suite in the Level 2 Vegetation Earth SystemData Record (VESDR)Product

Parametername UnitsResolution Comments

Temporal Spatial NormalizedDifferenceVegetationIndex(NDVI) none 65to110min 10km

differencebetweenBidirectionalReflectanceFactor(BRF)at779.5nmand680nmnormalizedbytheirsum

FractionvegetationabsorbedPhotosyntheticallyActiveRadiation(FPAR)

fraction 65to110min 10kmfractionofphotosyntheticallyactiveradiation(400–700nm)absorbedbyvegetation

LeafAreaIndex(LAI)!"#$%&'

!()*+%,' 65to110min 10km

one-sidedgreenleafareaperunitgroundareainbroadleafcanopiesandtheprojectedneedleareainconiferouscanopies

SunlitLeafAreaIndex(SLAI)𝑚./01234

𝑚567/084 65to110min 10km

one-sidedsunlitgreenleafareaperunitgroundareainbroadleafcanopiesandtheprojectedsunlitneedleareainconiferouscanopies

PrecisionofLeafAreaIndex(Dlai)

𝑚91:034

𝑚57/084 65to110min 10km retrievaldispersionofLAI

DirectionalAreaScatteringFactor(DASF) none 65to110min 10km

EstimateofCanopyBidirectionalReflectanceFactorasifthefoliagedoesnotabsorbradiation

QualityAssessmentvariable(QA_VESDR) none 65to110min 10km OverallqualityoftheVESDR

parametersWiththeexceptionofLAI,allVESDRparametersvarywiththesun-sensorgeometry.TheVESDRfilealsoincludesSolarZenithAngle(SZA),SolarAzimuthalAngle(SAA),ViewZenith(VZA)andAzimuthal(VAA)anglesatthesametemporalandspatialresolutions(Sect.3).The DSCOVR EPIC Science Algorithm Team also provides two ancillary science dataproducts,namely,10kmLandCoverTypeandDistributionofLandCoverTypeswithin10kmEPIC pixel. The products were derived from 500m MODIS land cover type 3 product(MCDLCHKM), which was generated from 2008, 2009 and 2010 land cover products(MCD12Q1,v051).TheancillarydatasetsaresummarizedinTable2.All products are projected on 10 km sinusoidal (SIN) grid and written in the standardHierarchical Data Format 5 (HDF5) using HDF-defined data models(http://www.hdfgroup.org/HDF5/). The EPIC VESDR and ancillary data products arepublicly available from the NASA Langley Atmospheric Science Data Center (https://eosweb.larc.nasa.gov/project/dscovr/dscovr_table).

3

Table2:Ancillarysciencedataproductderivedform500mMODISlandcovertype3product

Parametername Units Resolution CommentsTemporal SpatialLandCoverType none static 10km 10kmSINLandCovertype

LandCoverTypeDistribution none static 10km Distributionoflandcovertypeswithin10kmEPICpixel

1.2.Productmaturitylevel.DefinitionsofproductmaturitylevelsdevelopedbytheMISRteamareadopted(https://www-misr.jpl.nasa.gov/getData/maturityLevels/).TheDSCOVREPICVESDRproductisreleasedatProvisionalqualitylevel,i.e.,o Incrementalimprovementsarestilloccurring.Obviousartifactsorblundersobserved

inprereleaseproducthavebeenidentifiedandeitherminimizedordocumentedo Generalresearchcommunityisencouragedtoparticipateinthequalityassessmentand

validation, but need to be aware that product validation and quality assessment areongoing

o Parametermaybeusedinpublicationsaslongasprovisionalqualityisindicatedbytheauthors.Usersareurgedtocontactscienceteamrepresentativespriortouseofthedatainpublications,andtorecommendmembersoftheinstrumentteamsasreviewers

o TheDataQualitySummarystatesestimateduncertaintieso Maybereplacedinthearchivewhenanupgradedproductbecomesavailable,butshould

bereproducibleupondemandDSCOVREPICdataproductsbegin inaprovisionalstate,andadvancethroughaseriesofmaturitylevels,fromProvisionaltoValidatedstatus,i.e.fromadevelopmentalstatustoascientificallyprovenstatus.1.3. DSCOVR EPIC documents. Project documents are available athttps://eosweb.larc.nasa.gov/project/dscovr/dscovr_table.DSCOVREPICpublicationscanbe found at https://epic.gsfc.nasa.gov/science/pubs. The VESDR theoretical basis wasdocumentsin[1] Yang,B.,Knyazikhin,Y.,Mõttus,M.,Rautiainen,M.,Stenberg,P.,Yan,L.,Chen,C.,Yan,K.,

Choi, S.,Park,T.,&Myneni,R.B. (2017).Estimationof leaf area index and its sunlitportion fromDSCOVREPIC data: Theoretical basis.Remote Sensing of Environment,198,69-84.doi:/10.1016/j.rse.2017.05.033

AnoverviewoftheDSCOVREPICprojectisdocumentedin[2] Marshak,A.,Herman,J.,Szabo,A.,Blank,K.,Cede,A.,Carn,S.,Geogdzhayev,I.,Huang,

D.,Huang,L.-K.,Knyazikhin,Y.,Kowalewski,M.,Krotkov,N.,Lyapustin,A.,McPeters,R.,Torres,O.,&Yang,Y.EarthObservationsfromDSCOVR/EPICInstrument.BulletinoftheAmericanMeteorologicalSociety,doi:/10.1175/BAMS-D-17-0223.1.

4

TheDirectionalAreaScatteringFactor(DASF)isanewstructuralparameterthatestimatesthecanopyBRFiftheleavesdonotabsorbradiation.Itsdefinitionandanalysisofitsvalueforremotesensingofleafbiochemistrycanbefoundin[3] Knyazikhin,Y.,Schull,M.A.,Stenberg,P.,Mõttus,M.,Rautiainen,M.,Yang,Y.,Marshak,

A.,LatorreCarmona,P.,Kaufmann,R.K.,Lewis,P.,Disney,M.I.,Vanderbilt,V.,Davis,A.B.,Baret,F.,Jacquemoud,S.,Lyapustin,A.,&Myneni,R.B.(2013).Hyperspectralremotesensingoffoliarnitrogencontent.ProceedingsoftheNationalAcademyofSciences,110,E185-E192

1.4.Revisions.Thisisthefirstversionofthedocument.Itcanbedownloaded,distributed,andcited.RevisionsoftheScienceDataProductGuidewillbedetailedinthissection.2.EXPERIMENTOVERVIEWTheDeepSpaceClimateObservatory(DSCOVR)missionisamultiagency(NationalOceanicandAtmosphericAdministration[NOAA],U.S.AirForce,andNASA)missionlaunchedfromCapeCanaveral,FloridaonFebruary11,2015withtheprimarygoalofmakinguniquespaceweathermeasurementsfromthefirstSun-EarthLagrangepoint(L1).TheL1pointisonthedirect line between Earth and the Sun located 1.5million km sunward from Earth. ThespacecraftisorbitingthispointinasixmonthLissajousorbitwithaSun-Earth-View(SEV)anglevaryingbetween4.5oand11.5o.TheprimaryscienceobjectiveoftheDSCOVRmissionistoprovidesolarwindthermalplasmaandmagneticfieldmeasurementstoenablespaceweatherforecastingbyNOAA.The DSCOVR hosts NASA Earth-Observing Instrument, the Earth Polychromatic ImagingCamera(EPIC).TheEPICprovidesmeasurementsoftheradiationreflectedbyEarthintenwavelengthsandimagesofthesunlitsideofEarthforscienceapplications.2.1.EPICinstrumentcharacteristics.TheEPICinstrumentcollectsmultispectraldataoftheEarthintenwavelengths.ThespectralbandcharacteristicsaresummarizedinTable3.Table3:EPICspectralbandcomposition

Wavelength,nm FWHM,nm NominalProduct

317.5±0.1 1±0.2 Ozon325±0.1 2±0.2 Ozon340±0.3 3±0.6 Ozon,Aerosols,Clouds388±0.3 3±0.6 Aerosols,Clouds443±1 3±0.6 Aerosols551±1 3±0.6 Aerosols,Vegetation

680±0.2 2±0.4 Aerosols,Vegetation,Clouds,O2B-BandReference

687.75±0.2 0.8±0.2 O2B-BandCloudHeight764±0.2 1±0.2 O2A-BandCloudHeight,AerosolHeight779±0.3 2±0.4 O2A-BandReference,Vegetation

5

2.2. Rationale for the DSCOVR EPIC VESDR product. Fraction vegetation absorbedPhotosyntheticallyActiveRadiation (FPAR), Leaf Area Index (LAI), its sunlit counterpart(SLAI), and Normalized Difference Vegetation Index (NDVI are useful for (a)monitoringvariabilityandchangeinglobalvegetationduetoclimateandanthropogenicinfluences,(b)modelingclimate,carbonandwatercycles,and(c) improving forecastingofnearsurfaceweather.TheDirectionalAreaScatteringFactorprovidesinformationcriticaltoaccountingfor structural contributions tomeasurements of leaf biochemistry from remote sensing.WhereasLAIisastandardproductofmanysatellitemissions,globaldiurnalcoursesofFPAR,NDVI,SLAIandDASFarenewsatellitederivedproducts.3.VESDRSUN-SENSORGEOMETRYThesun-sensorgeometryisexpressedinaright-handedcoordinatesysteminwhichtheZ-axis(shownas“+Z”inFig.1)isalignedwiththenormaltothesurfacereferenceellipsoid(definedbytheWorldGeodeticSystem1984,WGS84),andpointstowardthecenteroftheEarth.TheX-axisisalignedwithagreatcircleandpointstowardthenorthpole.TheY-axisisorthogonaltobothofthem.

Figure 1. Right-handed coordinate system inwhichtheZ-axis(shownas“+Z”)pointstowardthecenter of the Earth. The X-axis and Y-axis pointtoward the North and East, respectively. Thedirection(unitvector)Ω<hasanazimuthalangle,𝜑< ,measuredclockwisefromthelocalnorthvector(X)totheprojectionofΩ<ontotheXYplane,andapolarangle,𝜃<,withrespecttothe+Zaxis.

TheSunandsensorpositionsarerepresentedbyunitvectorsΩ<andΩdirecteddownwardfromtheSuntotarget(i.e.,pointattheEarthsurface)andfromsensortotarget,respectively.Their polar (θ< and θ) and azimuthal (φ< andφ) angles are given in the right-handedcoordinatesystem(Fig.1).Theirrangesarebetween0and90°(polarangles)andbetween0 and 360° (azimuthal angles). This coordinate system is inherited from the upstreamDSCOVEREPICL2MAIACsurfacereflectanceproduct,whichisinputtotheVESDRretrievalalgorithm. In this coordinate system the Solar ZenithAngle (SZA, the angle between thetarget-to-Sundirection,−Ω<,andthe–Zaxis)coincideswiththepolarangleofΩ<,i.e.,θ< =𝑆𝑍𝐴.TheEarth-observinggeometryoftheEPICinstrumentischaracterizedbyanearlyconstantphaseangle(theanglebetweendirectionstotheSunandtothesensor)between4.5oand11.5°.Thephaseangle,𝛾,canbecalculatedas𝛾 = acos(cos𝜃 cos 𝜃< + sin 𝜃 sin 𝜃< cos(𝜑 − 𝜑<)). (1)

+Z

X (North)

Y (East)

-Z

j0

q0W0

q0

6

4.PRODUCTTILINGTheVESDRandancillarysciencedataproductsareprojectedon10kmsinusoidalgrid.Theglobe is divided into 4 horizontal tiles along the east-west, and 2 vertical tiles along thenorth-southaxes(Fig.2).Eachtileisidentifiedbyitshorizontal(from0to3)andvertical(from0to1)coordinates,e.g.,tile01.Dimensionofonetileis1000x1002.

Figure2.10kmSINDSCOVREPIClandcovertype.Theglobeisdividedinto8equaltiles.Eachtileisidentifiedbyitshorizontal(from0to3)andvertical(from0to1)coordinates.TheVESDRandancillarysciencedataproductsusethistilingstructure.5.LEVEL2VESDRPRODUCT5.1.VESDRproductfilenameThefilenamecontainingVESDRparametersisDSCOVR_EPIC_L2_VESDR_V1_YYYYMMDDHHMMSS_V2.h5Here V1 and V2 are versions of the VESDR product and L2B TOA reflectance data,respectively. Current versions areV1=01 andV2=02.YYYYMMDDHHMMSSsignifies dateand GMT time of EPIC image acquisition. For example, fileDSCOVR_EPIC_L2_VESDR_01_20160823141930_02.h5 contains VESDR parameters for anEPICimageacquiredonAugust23,2016(20160823)at14h10m30sGMT(141930).5.2.HDFfilestructureTheVESDRproductisdistributedasstandardHierarchicalDataFormat5(HDF5)file.Thedata are compressed using the lossless gzip option provided by the HDF5 FORTRANApplication Programming Interface (API). Compression level is 4. On average L2 VESDRproductisabout20-22megabytes(MB).

7

In theHDF5 file,data aregroupedby tiles.Eachgroupcontainsgeophysicalparameters,associatedqualityassessmentvariables(QA_VESDRandDlai)andsun-sensorgeometry.TherootgroupdirectorycontainsasetofattributesthatdescribesthecontentoftheHDF5file.Figure3illustratesasnapshotofthelayoutofaL2VESDRproduct.

Figure3.StructureoftheL2VESDRproduct. Data are grouped by tiles.Each group contains geophysicalparamet-ers, associated qualityassessmentvariablesandsun-sensorgeometry. The root attributedirectory provides generalinformation about the VESDRproduct.

5.3.RootattributesRootattributesincludedateandtimeoftheEPICimageacquisition,fillvalues,parameter’svalid ranges,mapprojection, scale factors and listof tiles present in the file. Details aresummarizedinTable4.Table4:L2VESDRrootattributes

Attributename Value,range Type DescriptionDate,YYYYMMDD 20160613-currentdate 32bitinteger dateofEPICimageacquisitionDate.GMT,hHMMSS 00000-235959 32bitinteger GMTofEPICimageacquisition

Fill_value_VESDR -9999 16bitinteger VESDRparameterwasnotgenerated

Fill_value_land -9998 16bitinteger non-vegetatedpixel

Fill_value_map -9997 16bitinteger outofmapboundary(“blackarea”inFig.2)

Fpar/ndvi/dasfvalidrange 0-1000 string validrangeofFPAR,NDVIandDASFLAI/SLAI/Dlaivalidrange 0-6850 string validrangeofLAI,SLAIanddLAI

MaxSZAthreshold 74.0 32bitfloatingpoint

VESDRalgorithmdoesnotprocesspixeliftheSZAexceedsMax_SZA_threshold

Mapprojection 10kmSIN,centermeridianis0 string VESDRparametersareprojectedon

10kmsinusoidalgrid

Groups

Datasets

Rootattributes

8

Scale_factor_VESDR 0.001 32bitfloatingpoint

VESDRparametershouldbemultipliedbythescalefactortoconvertitsDNvaluetophysicalvalue

Scale_factor_angle 1.0 32bitfloatingpoint

Sun-sensorgeometryparametersshouldbemultipliedbythescalefactortogettheirphysicalvalue

Totaltilespresent 1-8 8bitinteger TotalnumberoftilespresentintheVESDRfile

tile00_present 0,1 8bitinteger Indicatesiftile00present(value=1)indataset

tile01_present 0,1 8bitinteger Indicatesiftile01present(value=1)indataset

tile02_present 0,1 8bitinteger Indicatesiftile02present(value=1)indataset

tile03_present 0,1 8bitinteger Indicatesiftile03present(value=1)indataset

tile10_present 0,1 8bitinteger Indicatesiftile10present(value=1)indataet

tile11_present 0,1 8bitinteger Indicatesiftile11present(value=1)indataset

tile12_present 0,1 8bitinteger Indicatesiftile12present(value=1)indataet

tile13_present 0,1 8bitinteger Indicatesiftile13present(value=1)indataset

5.4.DatasetsEach group contains geophysical parameters, associated quality assessment variables(QA_VESDRandDlai)andsun-sensorgeometry.DescriptionofthedatasetsisgiveninTable5.Table5:L2VESDRdatasets

Nameofdataset

Validrange Datatype Description

01_LAI 0-6850 16bitinteger LeafAreaIndex02_SLAI 0-6850 16bitinteger SunlitLeafAreaIndex

03_FPAR 0-1000 16bitinteger fractionofphotosyntheticallyactiveradiation(400–700nm)absorbedbyvegetation

04_Dlai 0-6850 16bitinteger PrecisionofLeafAreaIndex05_NDVI 0-1000 16bitinteger NormalizedDifferenceVegetationIndex06_QA_VESDR 0-767 16bitinteger QualityAssessmentvariable.Seesection5.5

07_SZA 0-90 32bitfloatingpoint Polarangle(inDEG)oftheSun-to-targetdirectionasdefinedinSect.3

08_VZA 0-90 32bitfloatingpoint Polarangle(inDEG)ofthesensor-to-targetdirectionasdefinedinSect.3

09_SAA 0-360 32bitfloatingpoint Azimuthalangle(inDEG)oftheSun-to-targetdirectionasdefinedinSect.3

10_VAA 0-360 32bitfloatingpoint Azimuthalangle(inDEG)ofthesensor-to-targetdirectionasdefinedinSect.3

11_DASF 0-1000 16bitinteger EstimateofCanopyBidirectionalReflectanceFactorasifthefoliagedoesnotabsorbradiation

9

5.5.Qualityassessmentdataset5.5.1. Information content of QA dataset. Quality assessment variable, 06_QA_VESDR,includes quality control information on VESDR algorithm performance (bits 0 to 5) andStatus_QA (bits 6 to9). The latter is provided by the upstreamDSCOVREPIC L2MAIACsurface reflectance product. The DSCOVR EPIC MAIAC product is input to the VESDRretrieval technique. 06_QA_VESDR therefore provides information about quality of bothinputtotheVESDRalgorithmandtheVESDRalgorithmoutput.Figure4showsstructureof06_QA_VESDR.DetailsaregiveninTable6.

Figure4.Informationcontentof06_QA_VESDR

Table5:ValuesofQA_VESDRQAname Bits Binary

valueDecimalvalue

Description

VESDRalgorithmpath

0-1

00 0 VESDRparametersproduced.Nosaturation01 1 VESDRparametersproducedunderasaturationcondition10 2 VESDRalgorithmfailedtogenerateparameters11 3 VESDRparameterswerenotproduced.Otherreasons.Seebits2-5

Inputqualitytest 2-3

00 0 Inputqualitytestpassed01 1 Inputqualitytestfailed

10 2 InputqualitytestwasnotperformedbecauseBRFatNIRand/orGreenspectralbandswerenotavailable

11 3 VESDRparameterswerenotproducedbecausepixelwasnot-vegetatedoroutofmap.Bits0-5aresetto1inthiscase

Inputavailability 4

0 0 BRFatNIRandRedspectralbandswereavailable

1 1 ParameterswerenotproducedbecauseBRFsatNIRand/orRedspectralbandswerenotavailable.

SZA 5 0 0 SZAisbetween0oandmaxSZAthreshod.Seerootattributes.1 1 SZAisoutsideoftheacceptablerange.Parametersnotproduced

Status_QA 6-9

0000 0 Noclouds,CM_CLEAR_WATER0001 1 Noclouds,CM_CLEAR_WATERSED0010 2 1neighborcloud0011 3 >1neighborclouds0100 4 noretrieval(cloudy,orwhatever)0101 5 definitionisnotprovided0110 6 forH>3.5km,noretrieval0111 7 definitionisnotprovided0100 8 sunglint1001 9 land-watermisclassified1010 10 CoxMunktoohigh1011 11 infonotavailable

VESDR algorithm path00: produced without sat01: produced, saturation10: algorithm fails11: not produced

Input quality test00: passed01: failed10: not performed11: non-veg or outside map

Input0: available1: not available

SZA 0: within range1: out of range

Status_QA, copied from the MAIAC BRF product

10

5.5.2.Saturation,RetrievalIndexandinputqualitytest.Inthecaseofdensecanopies,thereflectancessaturateandthereforeareweaklysensitivetochanges incanopyproperties.The reliability of parameters retrieved under the condition of saturation is low. Suchretrievals are flaggedby settingdecimalvalueof theVESDR_algorithm_path to1 (binaryvalue=’01’). The retrieval index,RI, is the percentage of pixelswith valid BRF forwhich the VESDRalgorithm produced a retrieval. The index characterizes the spatial coverage of thegeophysicalparameters.Thisimportantcharacteristicofthealgorithmperformancecanbecalculatedas

𝑅𝐼 =𝑁(bits01 =\ 00\orbits01 =\ 01\)

𝑁(bit4 =\ 0\) . (2)

HerethenumeratorrepresentsnumberofpixelsforwhichtheVESDR_algorithm_pathis0or1.Thedenominatoristhenumberofpixelsforwhichavalueofbit4is0. Forvegetatedpixelsatweaklyabsorbingwavelengths,theBRFtoleafalbedoratioislinearlyrelatedtoBRF,i.e.,`abcdc

= 𝑝𝐵𝑅𝐹h + 𝑅, (3)wheretheslope,p,andintercept,R,aretherecollisionprobabilityandescapefactor.WeuseBRF at green and NIR EPIC bands to estimate the slope, p, of a line passing pointsi`abjkllmdjkllm

, 𝐵𝑅𝐹opqqrsandt`abuvwduvw

, 𝐵𝑅𝐹xyazonthe`abdvs𝐵𝑅𝐹plane.Itsvalueisgivenby

𝑝 ={w|uvw}uvw

~{w|jkllm}jkllm

`abuvw~`abjkllm. (4)

Here𝜔hrepresentsafixedleafalbedoatNIRandgreenspectralbands.Itsvaluesatthesebandsaresetto0.4898(green)and0.9789(NIR)intheVESDRoperationalalgorithm.OuranalysessuggestthatEq.(4)takesvaluesbetween0and1onlyforvegetatedsurfaces.ForBRFatgreenandNIRspectralbandsovernon-vegetatedland,waterorcloud-contaminatedpixels,Eq.(4)generatesvaluesoutsideofthe0to1range.Thispropertyunderliestheinputqualitytest:bits2-3aresetto‘00’ifpisbetween0and1,andto‘01’,otherwise.TheVESDRalgorithmprocessespixelsirrespectiveofthetestresult.Theinput_quality_testQAisjustawarningthatVESDRparameterswereretrievedusinginputBRFofsuspiciousquality.Figure 4 shows distribution of α = atan(𝑝), 0° ≤ 𝛼 ≤ 180° derived from EPIC L1B TOAreflectancedata.Onecanseethatvaluesofαcorrespondingtocloudfreeland,vegetation,oceanandcloudcontaminatedpixelstendtooccupydifferentspaceswithinthe0oto180ointerval.

11

Figure4.LeftpanelshowsanEPICRGBimagetakenonAugust23,2016at15:24:58GMT.Eq.(4)wasappliedtoeachimagepixel.Theslopepwasconvertedtoangle,α = atan(𝑝),𝛼 ∈ [0°, 180°],betweenthelineandBRFaxis.Rightpanel showsdistributionofα over theEPIC image. Itsvaluescorresponding tocloud free land,vegetation,oceanandcloudcontaminatedpixelstendtooccupydifferentspaceswithinthe0oto180ointerval.6.ANCILLARYSCIENCEDATAPRODUCTSAlandcovermapisanimportantancillarydatalayerusedbytheVESDRretrievalalgorithm.The global classification of canopy structural types utilized in the Collection 6 MODISLAI/FPAR algorithm is adopted. Global vegetation is stratified into eight canopyarchitecturaltypes,orbiomes.TheeightbiomesareGrassesandCerealCrops(B1),Shrubs(B2), Broadleaf Crops (B3), Savannas (B4), EvergreenBroadleaf Forests (B5),DeciduousBroadleaf Forests (B6), Evergreen Needle Leaf Forests (B7) and Deciduous Needle LeafForests(B8).TheVESDRancillarysciencedataproductsinclude10kmLandCoverTypeandDistributionofLandCoverTypeswithin10kmEPICpixel.TheseproductswerederivedfromtheMODIS8-biome SIN 500 m resolution land cover type 3 product (MCDLCHKM), which wasgeneratedfrom2008,2009and2010MODISlandcoverproducts(MCD12Q1,v051).6.1.DSCOVREPIClandcovertypeTheMODISLandCoverProductisprojectedon500msinusoidal(SIN)grid.A10kmEPICSINgridpixelthereforecontainsabout400MODISpixelswithknownlandcovertypes.TheEPIC landcover type is assignedbasedon thedominant landcover fraction. If thereareseverallandcovertypeswithequalfrequency,biometypewithhighestbiomenumberBNistaken as the EPIC land cover type. For example, if B5 (DeciduousBroadleaf Forests), B4(Shrubs)andB1(GrassesandCerealCrops)occupy40%,40%and20%ofthepixelarea,thenB5isassignedtotheEPIClandcovertype.ThemostfrequentlandcovertypenumbersarealsostoredintheDSCOVREPIClandcoverfile.Figure2shows10kmSINDSCOVREPIClandcovertype.DSCOVREPIClandcoverproductfilenameisDSCOVR_EPIC_ANC_LCTYPE_MCD12Q1_51.h5.

Cloudfreeland

Cloudcontaminatedpixels

Cloudfreeocean

Cloudfreevegetation

12

6.1.1.HDFfilestructure.IntheHDF5file,dataaregroupedbytiles.Eachrootgroupcontainstwodatasets,“Land_Cover_Type_3”and“Multi_Land_Cover_Types_presented,”aswellasatile group “Geolocation” with two datasets, “Latitude” and “Longitude.” The root groupdirectorycontainsasetofattributesthatdescribesthecontentoftheHDF5file.

Figure5.Structureoftheancillary10kmEPICLandCoverTypeProduct

6.1.2.Rootattributes.RootattributesincludelandcovertypeIDsandassociatedvalues,fillvalues, parameter valid ranges and short description of the“Multi_Land_Cover_Types_presented”dataset.DetailsaresummarizedinTable6.Table6:LandCovertyperootattributes

Attributename Value Type Description00_Water 0 8bitinteger Pixelisclassifiedaswater

01_Grasses/CerealCrops 1 8bitinteger PixelisclassifiedasB1:GrassesandCerealCrops

02_Shrubs 2 8bitinteger PixelisclassifiedasB2:Shrubs03_BroadleafCrops 3 8bitinteger PixelisclassifiedasB3:BroadleafCrops04_Savannas 4 8bitinteger PixelisclassifiedasB4:Savannas

O5_EvergreenBroadleafForests 5 8bitinteger PixelisclassifiedasB5:EvergreenBroadleafForests

06_DeciduousBroadleafForests 6 8bitinteger PixelisclassifiedasB6:DeciduousBroadleafForests

07_EvergreenNeedleLeafForests 7 8bitinteger PixelisclassifiedasB7:EvergreenNeedleLeafForests

08_DeciduousNeedleLeafForests 8 8bitinteger PixelisclassifiedasB8:Deciduous

NeedleLeafForests09_Non-vegetatedland 9 8bitinteger Pixelisclassifiedasnon-vegetatedland

Rootattributes

Rootgroups

Tilegroup

Tilegroupdatasets

Rootgroupdatasets

13

10_Urbanarea 10 8bitinteger Pixelisclassifiedasurbanarea11_Fillvalue 127 8bitinteger Outofmappixel(“Blackpixels”inFig.2)

Landcoverunits biomenumber string LandcovertypeID

Landcovervalidrange 0-11 string Validrangeinthe“Land_Cover_Type_3”dataset

Latvalidrange -9000-+9000 string Validrangeinthe“Latitude”dataset

Lat/Lonfillvalue 25500 16bitinteger

Fillvalueinthe“Latitude”and“Longitude”datasets

Lat/Lonscalefactor 0.0132bitfloatingpoint

LatitudeandLongitudeshouldbemultipliedbythescalefactortoconverttheirDNvaluestophysicalvalues

Lat/Lonunits degree string Unitsoflatitudeandlongitude

Lonvalidrange -18000-+18000 string Validrangeinthe“Longitude”dataset

Multilandcovertypespresenteddescription description string

Descriptionofthe“Multi_land_cover_typespresented”dataset.SeeSect.6.1.4.

Multilandcovertypespresentedfillvalue -9999 string Fillvalueinthe“Multi_land_cover_types

presented”dataset

Multilandcovertypespresentedvalidrange 0-4095 string

Validrangeinthe“Multi_land_cover_typespresented”dataset

6.1.3. Datasets. Each root group contains two data sets, “Land_Cover_Type_3” and“Multi_Land_Cover_Types_presented,”aswellastilegroup“Geolocation”withtwodatasets,“Latitude”and“Longitude”(Fig.5).DescriptionofthedatasetsaregiveninTable7.Table7:L2LandCovertypedatasetsNameofdataset Validrange Datatype DescriptionLand_Cover_Type_3 0-11 8bitinteger LandcovertypeMultilandcovertypespresented 0-4095 8bitinteger Informationaboutmultiplelandcovertypeswith

equalfrequency.SeeSect.6.1.4Latitude -9000-+9000 8bitinteger LatitudeLongitude -18000-+18000 8bitinteger Longitude6.1.4.Multilandcovertypespresented.Thisdatasetprovidesinformationaboutdominantlandcovertypesin10kmEPICpixels.Thisinformationisstoredin16-bitintegernumber.Bits0to10representlandcovertype(Fig.6),withbitvalue1indicatingdominantlandcovertype. For example, if water, B2 (Shrubs), B4 (Savannas) and B5 (Evergreen BroadleafForests) represent 10%, 30%, 30% and 30% of the pixel area, then Multi_Land_CoverTypes_Presentedis’110100’=52.Bit11withbitvalue1indicatesfillvalue,i.e.,landcovertype was not identified. In this case Multi_Land_Cover Types_Presented is’100000000000’=4095.

Figure 6. Structure of Multi_Land_CoverTypes_Presented

Bit 15 Bit 14 bit 13 bit 12 bit 11 bit 10 bit 9 bit 8 bit 7 bit 6 bit 5 bit 4 bit 3 bit 2 bit 1 bit 0

0 0 0 0 Fill value urban Non-vegetated B8 B7 B6 B5 B4 B3 B2 B1 water

14

6.2.Distributionoflandcovertypes. Distribution of land cover types within 10 km EPIC pixel is defined as

𝐿𝐶� = 100%𝑁�𝑁 .

(5)

Herei(i=0,1,2,…,10)representslandcovertypeID,NiisthenumberoftheithlandcovertypeandNisthetotalnumberofpixels;∑ 𝐿𝐶��<

< = 100%.SetsoftheMODIS500mlandcovertype3productwithin10kmEPICpixelwasusedtoderivethisdistribution.DSCOVREPIClandcoverproductfilenameisDSCOVR_EPIC_ANC_LCDIST_MCD12Q1_51.h5.6.2.1.HDFfilestructure.IntheHDF5file,dataaregroupedbytiles.Eachgroupcontainsdistributionoflandcovertypewithin10kmEPICpixel.TherootgroupdirectorycontainsasetofattributesthatdescribesthecontentoftheHDF5file.Figure7illustratesasnapshotoftheHDFViewlayoutofSCOVR_EPIC_LCDIST_MCD12Q1_51.h5file.

Figure7.Structureoftheancillary10kmSINEPICLandCoverTypeDistribution

6.2.2.Rootattributes.Rootattributesincludefillvalue,scalefactor,unitsanddistributionvalidrange.DetailsaresummarizedinTable8.

Groups

Rootattributes

Datasets

15

Table8:LandcovertypedistributionrootattributesAttributename Value Type Description

Histogramfillvalue -9999 16bitinteger Distributionfillvalue

Histogramscalefactor 0.01 32bitfloatingpoint

DistributiondatasetshouldbemultipliedbythescalefactortoconvertitsDNvaluetophysicalvalue

Histogramunits % string Unitsofthedistribution

Histogramvalidrange 0-100 string Validrangeaftermultiplyingthedistributionbythescalefactor

6.2.3.Datasets.Eachgroupcontainsdistributionoflandcovertype.DetailsaresummarizedinTable9.Table9:Landcovertypedistributiondatasets

Nameofdataset Value Type Description00_Water 0-10000 16bitinteger PercentageofwaterinEPICpixel

01_Grasses/CerealCrops 0-10000 16bitinteger PercentageofB1:GrassesandCerealCropsinEPICpixel

02_Shrubs 0-10000 16bitinteger PercentageofB2:ShrubsinEPICpixel

03_BroadleafCrops 0-10000 16bitinteger PercentageofB3:BroadleafCropsinEPICpixel

04_Savannas 0-10000 16bitinteger PercentageofB4:SavannasinEPICpixel

O5_EvergreenBroadleafForests 0-10000 16bitinteger PercentageofB5:EvergreenBroadleafForestsinEPICpixel

06_DeciduousBroadleafForests 0-10000 16bitinteger PercentageofB6:DeciduousBroadleafForestsinEPICpixel

07_EvergreenNeedleLeafForests 0-10000 16bitinteger PercentageofB7:EvergreenNeedleLeafForestsinEPICpixel

08_DeciduousNeedleLeafForests 0-10000 16bitinteger PercentageofB8:DeciduousNeedleLeafForestsinEPICpixel

09_Non-vegetatedland 0-10000 16bitinteger Percentageofnon-vegetatedlandinEPICpixel

10_Urbanarea 0-10000 16bitinteger PercentageofurbanareainEPICpixel11_Fillvalue 0-10000 16bitinteger PercentageoffillvalueinEPICpixel7.KNOWNISSUESThere are still residual issues in Level 1B data that affect the geolocation accuracy. Thisincludeserrorswiththestar-trackerpointing,accuracyofthetelescopeopticalmodel,imagetimestamps,andeffectsofatmosphericrefraction.Geolocationuncertainties impactbothatmosphericallycorrectedsurfacereflectanceandtheVESDRproducts.Work iscurrentlyunderway that treats these additional corrections to further improve science productsbeyondthebasicrequirements.8.EXAMPLES1.ObtainingnewinformationonvegetationpropertiesfromtheVESDRproduct.LAIcan vary significantlywith SLAI unaltered. This happens because the amount of shadedleaves can increase with SLAI unchanged. The goal of this section is to obtain canopy

16

interceptance, 𝑖<, direct transmittance, 𝑡<, Clumping Index (CI) andFractionalVegetationCover(FVC)fromLAI,SLAIandSZA.In spite of a weak correlation between LAI and SLAI due to the above feature, theseparameterssatisfythefollowingequation(publ.[1]insect.1.3)

𝑆𝐹 =1 − exp(−𝜏)

𝜏 . (E1)

HereSF=SLAI/LAIistheSunlitFraction(SF)ofleafarea,𝜏 = 𝐺 ∙ 𝐿𝐴𝐼 ∙ 𝐶𝐼/𝜇representstheopticalpaththroughthevegetationlayer,GandCIsignifythegeometryfactorandclumpingIndex (CI), and 𝜇 = cos 𝑆𝑍𝐴. The exponent exp(−𝜏) is the direct transmittance 𝑡<. ThisfeatureallowsustoderiveCIandFVCusingLAI,SLAIandSZA.Problem1:findclumpingindex(CI)givenLAI,SLAIandSZA.Thiscanbedonebasedonthefollowingalgorithm

Ø ForagivenpixelcalculateSF=SLAI/LAI;Ø SolveEq.(E1)for𝜏;Ø EstimateCIas𝐶𝐼 = 2𝜏𝜇/𝐿𝐴𝐼.

ThefollowingfeatureoftheVESDRalgorithmisusedhere.ThealgorithmaccumulatesLAIand corresponding direct transmittances forwhich simulated and observed BRFs agreewithinmodel and observation uncertainties.Mean LAI is reported as retrieved LAI. Thealgorithm averages direct transmittance, 𝑡<̅ = mean𝑡<, and calculates SF as 𝑆𝐹 = (1 −𝑡<̅)/| ln 𝑡<̅ |. The SLAI is estimated as𝑆𝐹 ∙ 𝐿𝐴𝐼. Thus, the physically based Eq. (E1) is notviolatedinthealgorithm’sframework.ThistrickminimizesimpactoferrorsinLAIonSLAI.Indeed,iftheretrieved𝐿𝐴𝐼 = 𝑘𝐿𝐴𝐼�p�q,thenanerrorinSFisinverselyproportionaltok,i.e.,𝑆𝐹~𝑆𝐿𝐴𝐼/𝑘𝐿𝐴𝐼~𝑆𝐹�p�q/𝑘.Theproduct𝑆𝐹 ∙ 𝐿𝐴𝐼thereforetendstocanceltheerrork.IntheabovealgorithmonesolvesEq.(E1)using𝑆𝐹~𝑆𝐹�p�q/𝑘.Itssolutionis𝜏~𝜏�p�q𝑘.Theratio𝜏/𝐿𝐴𝐼~𝜏�p�q𝑘/𝑘𝐿𝐴𝐼�p�qthereforealsotendstominimizeimpactofvariationinLAIonCI.Problem2:Findcanopyinterceptance,𝑖<,anddirecttransmittance,𝑡<,given𝜏.ItdirectlyfollowsfromEq.(E1)that𝑖<isjusttheproductbetween𝜏andSF,i.e.,𝑖< = 𝑆𝐹 ∙ 𝜏.Thedirecttransmittanceis𝑡< = 1 − 𝑖<.Problem3:FindFVCgiven𝑡<.Obviously,𝐹𝑉𝐶 = 1 − 𝑡<

¡.VESDRparametersfor2016-08-23areusedtoillustratetheabovealgorithms.Thereare20imagesacquiredduringthisday.Weselectbiome5(EvergreenBroadleafForests)intile11,whichrepresentAmazonian rainforests.EPIC sees this forestwithinaGMT time intervalbetween12hand18h.FigureE1showsLAIimagesofBiome5inthetile11.

17

FigureE1.EPICLAIforGMT=12,14,16and18GMT.SouthAmericalocatedattheleftedgeoftheEPICimageat12:08GMT.SZAformostofthepixelsisabout650atthistime(Fig.E2).Geo-registrationerrorisveryhigh,whichsignificantlyimpactqualityofMAIACBRFdataandconsequentlyLAIretrievals.WedonotrecommendusingprovisionaldataforSZA>55o.Weselecteda5x5pixelareainAmazon,whichislocatedbetweentile11rows97and101andcolumnsbetween401and411,i.e.,area=/tile11/01_LAI(97:101,407:411).MeanSZAinthisareaat12:08GMTwas56.88o(std=0.126o)

FigureE2.DistributionofSZAforB5inthetile11at12:08GMT.Wefixeda5x5pixelareaforwhichSZA was around 56o. Mean SZA=56.88o,STD=0.126o, Coefficient of variation,100%STD/Mean,is0.2%

TableE1showsmeanSZA,LAIandSLAIoverthe5x5pixelarea.Correspondingcoefficientsofvariation(100%STD/Mean)areshowninparentheses.LAIshouldnotvarywithinthisareaduringaday.Thisisnotthecaseinourexample:variationinLAIisabout14%(orwithinabout0.7LAIunits),whichiscomparablewithC6MODISLAI(0.69LAIunits).Thiserrorisduemodelandobservationuncertainties.

18

AtagivenGMT(i.e.,forafixedSZA),SLAIshouldnotvarywithinourarea.VariationinthisparameterintheselectedareaatgivenSZAisabouttwicelowerthanvariationinLAI(e.g.,6.6%inSLAIvs13.1%inLAI).ThisconfirmsourtheoreticalresultsthatthealgorithmtendstominimizeimpactoferrorsinLAIonSLAIretrievals.CIandFVCshouldbeconstantwithinourarea.Theirvariationsareabout3.1%and3.7%(cf.with14%inLAI),respectively,whichcanbetreatedastheprecisionoftheseparameters.Thisisagoodvaluefortheproductprecision.Table E1. Mean LAI, SLAI, 𝜏, CI, 𝑡< and FVC over our 5x5 pixel area. Coefficients of variation,Var=100%std/mean,areshowninparentheses.Lastthreecolumnsshowdailymean,stdandcoefficientofvariationforSZAindependentvariables.GMT,hh:mm 12:08 13:14 14:19 15:24 16:30 17:35 18:41 Daily

mean STD Var,%

SZA(var,%) 56.88(0.2)

42.15(0.4)

29.03(0.5)

20.70(0.6)

23.28(0.6)

34.15(0.3)

48.15(0.3) n/a n/a n/a

LAI(var,%) 5.32(13.1)

5.99(3.9)

6.02(1.9)

6.34(2.8)

5.59(7.6)

5.38(8.0)

3.99(30.6) 5.52 0.77 14.0

SLAI(var,%) 1.51(6.6)

1.84(2.1)

2.10(0.4)

2.36(0.0)

2.22(2.1)

1.98(2.5)

1.56(12.4) n/a n/a n/a

SolutionofEq.(E1),𝝉 3.43 3.04 2.62 2.44 2.2 2.44 2.28 n/a n/a n/aClumpingIndex 0.704 0.752 0.761 0.719 0.723 0.750 0.763 0.739 0.023 3.13Directtransmittance 0.025 0.062 0.088 0.091 0.126 0.100 0.110 n/a n/a n/aFractionalVegetationCover 0.975 0.938 0.912 0.909 0.873 0.900 0.889 0.914 0.033 3.66