PracticalSession-WorkingwithairborneimagingspectrometerdataENVIDanClewley–NERC-ARF-DAN,PlymouthMarineLaboratory

1. IntroductionThispracticalsessionprovidesanintroductiontoworkingwithairborneimagingspectrometer(hyperspectral)datausingENVI.Bytheendofthispracticalyouwillbeableto:

1. OpenandviewairbornehyperspectraldatainENVI2. Performanempiricallinecorrectiontoconverttosurfacereflectance

2. AirborneData For this practical data from the Specim ASIAFenix instrument collected by the NERC AirborneResearchFacility(NERC-ARF1)willbeused.ThesedatawereflownovertheBarraxsiteinSpainon17th June2015aspartof theATSC summer school2. The lineused for thepractical is shown inFigure1.TheFenixinstrumentcoversvisibleandnearinfrared(VNIR)toshortwaveinfrared(SWIR)portionofthespectrum,wavelengths(380–2500nm)overamaximumof622bands.Forthispracticalasubsetofline5willbeused.Thislinewasflownatanaltitudeof1400mfromsoutheasttonorthwestat10:50(GMT).TheVNIRbandswerespectrallybinned,providingatotalof448bands,beforemaskingand439aftermasking.Thesedatahavebeenpre-processedbytheNERC-ARFDataAnalysisNode(NERC-ARF-DAN3)usingthefollowingsteps:1. RadiometricCalibration2. Masking out of bad pixels, including dropping completely masked out bands (at theVNIR/SWIRoverlap).3. GPSandInertialMeasurementUnit(IMU)processingandsynchronisation4. Mappingofdata5. Subsettingtoareaaroundtargets(toreduceprocessingtimeinpracticalsession)

1 https://www.bas.ac.uk/nerc-arf 2 https://www.nceo.ac.uk/training/atsc-training/ 3 https://nerc-arf-dan.pml.ac.uk/

Figure1:HyperspectrallineusedforpracticalsessionoverBarraxsiteinSpain.Subsetusedfor

practicalisshownincolour.

3. IntroductiontoENVI ENVI(EnvironmentforVisualisingImages)isacommercialsoftwarepackageforvisualisationandanalysis of remote sensing data. It has a number of features and algorithms targeted tohyperspectraldata.TherearetwointerfacesforENVI,thenewinterfaceandENVI‘classic’,forthispracticalsessionENVIclassicmodewillbeused.1. OpenENVIbyselecting“ENVIClassic”fromthestartmenu.2. Load the hyperspectral data into ENVI using “File”à “Open Image file” from themainmenu and selecting “f168053b_utm_wgs84N30_sub.bil”, you may need to navigate to thedirectorythepracticaldataarestoredfirst.3. Inthe“AvailableBandsList”windowrightclickon“f168053b_utm_wgs84N30_sub.bil”andselect“LoadTrueColour”.Thiswillopen3windowsasshowninFigure2:

• Scroll–Showsanoverviewoftheimage• Image–Showsasubsetandincludestheimagespecificmenubar• Zoom – Shows a zoomed in subset, the zoom can be adjusted using the – and +iconsinthelowerleftcorner.

4. Within the Imagewindowmenu select “Tools”à “Profiles”à “ZProfile (Spectrum)” tobringupthespectralprofilewindow.5. Withthespectralprofilewindowopenselectdifferentpixelstoviewtheprofile,trylookingat the spectra of the targets. As the image hasn’t been atmospherically corrected the profile

showsat-sensorradianceinunitsof𝑛𝑊𝑐𝑚%&𝑠𝑟%)𝑛𝑚%).Thered,greenandbluelinesshowthewavelengthsofthebandscurrentlydisplayed.6. AfteryouhavefinishedexploringtheimageyoucanclosetheSpectralProfilewindowbutkeeptheotherwindowsopen.

Figure2:ImagedisplaywindowsinENVI

4. EmpiricalLineCorrection 1. WiththeimageWindowstillopenfromtheprevioussectionselect“Tools”à“RegionofInterest”à“ROItool”fromtheimagemenu.2. TheROItoolisusedtoidentifythepixelsforeachtarget.

a) FromtheROIwindowselect“Window:”“Zoom”,soonlypixelsclickedintheZoomwindowarerecorded.

b) Select“ROI_Type”à“Point”,thisallowsyoutoselectindividualpixelsratherthandrawingpolygons,whichisbetterforsmallfeatures.

c) Select“Region#1”intheROIlistandrenameto“White”bydoubleclickingonthename,press“Enter”onceyouhaverenamed.

d) Inthezoomwindowselectpixelsfromthecentreofthewhitetargetbyclickingonthem as shown in Figure 3. It is important to select pixels towards the centre toavoid mixed pixels towards the edge of the target where the spectra will be amixtureofthetargetandthesurfaceithasbeenlaidouton.Ifyoumakeamistakeyoucanremoveapixelusingthemiddlemousebutton(pressonthescrollwheel).

e) AfteraROIhasbeendefinedforthewhitetarget,select“NewRegion”fromtheROIwindowandrepeatfortheblackandgreytargets.

Figure3:SelectingpixelsforWhitetargetusingROItool.

3. OnceanROIhasbeendefinedforeachtargetsavetheseusing“File”à“SaveROIs”fromtheROIwindow.Use“SelectAllItems”andusethe“Choose”buttontonavigatethedirectoryyouare working in. The recommended name for the ROI file is“f168053b_utm_wgs84N30_sub_targets.roi”4. YoucanviewthevariationinspectrawithineachofthetargetsbyclickingselectinganROIandclickingthe“Stats”button.Bydefaultitwillshowthemean(white)+/-1standarddeviation(green)andmin/max(red),youcanchangeusing“SelectPlot”.5. AftertheROIshavebeenidentifiedtheempiricallinecalibrationcanbeperformed.

a) Select “Spectral”à “Pre-processing”à “CalibrationUtilities”à “Empirical Line”à“ComputeFactorsandCalibrate”fromthemainmenu.

b) Select“f168053b_utm_wgs84N30_sub.bil”astheinputfile.c) For “DataSpectra” select “ImportSpectra” then 'Import”à “fromROI/EVG from

inputfile”.Selectall items(White,GreyandBlack),click“OK”followedby“SelectAll” and “Apply”to load into the Empirical Linedialogue, the importwindow canthenbeclosed.

d) For “Field Spectra” Select “Import Spectra” then “Import” à “from SpectralLibrary”.Use the “Open” dialogue to open a spectral library file.Navigate to thepractical directory and select “panels_avg_all_envi.sli”. This has been generatedusingtheaveragefieldspectrafromeachtarget.Selectall items(White,GreyandBlack) and click OK, then “Select All” and “Apply”to load into the Empirical Linedialogue,theimportwindowcanthenbeclosed.

e) Select“White”inthe“DataSpectra”boxand“White”inthe“FieldSpectra”boxandclick “Enter Pair” to identify them as amatching image spectra and field spectrapair.Repeatfortheothertwotargetsthenclick“OK”.

f) Intheoutputfilewindowselect“Chose”tonavigatetothepracticaldatadirectory

anduse “f168053b_utm_wgs84N30_sub_elc.bil” as theoutput file name, use thesame procedure and select “f168053b_utm_wgs84N30_sub_elc_cff.cff” for theCalibration Filename. The calibration file contains the results of the regressionbetween the imageand field spectraand canbeused toapply theempirical linecorrectiontolineswhichdon’tcontaintargets.ClickOKtostartrunning.

6. Once the Empirical Line Correction has finished a new file(f168053b_utm_wgs84N30_sub_elc.bil)willappearintheAvailableBandsListwindow.Rightclickonthisandselect“LoadTrueColourto<new>”.7. Withinthe Imagewindowmenuforthenewfileselect“Tools”à“Profiles”à“ZProfile(Spectrum)”tobringupthespectralprofilewindowandusethistoviewthespectra.Thevaluesarereflectanceandgofrom0–1,asshowninFigure4.Therearelikelytobenoisyregionsofthespectrumwherereflectanceisrecordedasbeingnegative,thesecanberemovedaspartofpostprocessingifrequired.8. Keepallthewindowsopenforthenextsection.

Figure4:Imagespectraofgrassafterempiricallinecorrection.

5. ExploringImageSpectra 1. Withinthespectralprofilewindowselect“Edit”à“PlotParameters”andchangetherangeoftheY-Axisfrom0to1topreventtherangefrommoving.2. Stillinthespectralprofilewindowdragtheredlineto~800nm,rightclickontheplotandselect “LoadNewRGB combination” to change thebandsdisplayed.Notice in this combinationhow vegetation (which has a high NIR reflectance) appears as bright red. Try differentcombinationsandnoticehowatdifferentbandreduceoremphasisethecontrastbetweencertainfeatures.3. Toexportthedatafromtheplotyoucanselect“File”à“SavePlotAs”à“ASCII..”fromtheplotwindow.Thiswillcreateatextfilewhichcanbeopenedinotherprograms,forexampleExcel.4. Asyoumovetheplottoolaroundtheimageyouwillnoticethespectravariesforsimilar

covers,todemonstratethisselect“Options”à“CollectSpectra”andmovethecursoraroundthegrasswhichthetargetsarelocatedon,theplotwillkeepaspectraforeachpointsoyoucanseethespreadofvalues.Untick“Options”à“CollectSpectra”andmovethecursortocleartheplot.5. Tozoominonaspecificareaofthespectrayoucanusethemiddlemousebutton(scrollwheel).

6. Furtherexercises Thispracticalhasprovidedabrief introductiontoairborneimagingspectrometerdata, includinghow to calibrate to surface reflectance using targets so each pixel is comparable withmeasurementsmadefromafieldspectrometer.Thereisalotmoreanalysiswhichcanbecarriedoutonthesedata, inparticularwhencombinedwith fieldmeasurements,a lotofwhichwillbespecifictoyourapplicationandtheresearchquestionsyouwishtoanswer.Ifyouhave ideasofwhatyouwould liketoachieveusingairbornedatabutaren’tquitesureofhow to go about it feel free to discusswithNERC-ARF and FSF.Outside theworkshop you cancontacttheNERC-ARFhelpdeskonnerc-arf-processing@pml.ac.ukSomefurtherexercisestotryinENVIduringthispracticalare:1. Spectrally resample the image to match a satellite sensor. Within ENVI use “Spectral”à”Spectral Resampling” and in the “Spectral Resampling Parameters” window select a Pre-DefinedFilterFunction (e.g., LandsatTM7).Compare thespectra fromthatof theoriginaldata.The resampled imagewith a reducednumber of bands is easier to view in aGIS program (e.g,ArcMap)2. Calculatevegetation indices suchas theNormalisedDifferencevegetation Index (NDVI4).Within ENVI this can be done using: “Spectral”à “Vegetation Analysis”à “Vegetation IndexCalculator”3. Using theROI tool,define regions forareasyouare interested in (e.g.,grass,water)andextractstatisticsforeachone,usingasimilarprocedureasusedforthetargets.

4 https://earthobservatory.nasa.gov/Features/MeasuringVegetation/