Open Collections

UBC Faculty Research and Publications

High temporal resolution NDVI phenology from micrometeorological radiation sensors Huemmrich, Karl F.; Black, T. Andrew; Jarvis, Paul G.; McCaughey, J. H.; Hall, Forrest G. Nov 30, 1999

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-Black_AGU_1999JD900164.pdf [ 860.1kB ]
JSON: 52383-1.0041908.json
JSON-LD: 52383-1.0041908-ld.json
RDF/XML (Pretty): 52383-1.0041908-rdf.xml
RDF/JSON: 52383-1.0041908-rdf.json
Turtle: 52383-1.0041908-turtle.txt
N-Triples: 52383-1.0041908-rdf-ntriples.txt
Original Record: 52383-1.0041908-source.json
Full Text

Full Text

JOURNAL  OF GEOPHYSICAL  RESEARCH,  VOL. 104, NO. D22, PAGES 27,935-27,944, NOVEMBER  27, 1999  High temporal resolution NDVI phenology from micrometeorological radiation sensors K. F. Huemmrich,• T. A. Black,2 P. G. Jarvis,• J. H. McCaughey,4 and F. G. Hall s Abstract. The boreal forest is a region characterizedby wide swingsin temperatureand light levelsover the courseof a year. This seasonalvariabilitystronglyeffectsthe vegetationof this biome. Normalizeddifferencevegetationindex (NDVI) valueswere observedat daily timescalesfor key land covertypesof the boreal forest, developinga more detailed descriptionof seasonalchangesin NDVI than could be producedfrom satellitedata. NDVI valueswere calculatedfrom tower-mountedphotosyntheticallyactive radiation(PAR) and globalsolarsensorsmeasuringboth incomingand reflectedradiation abovethe canopiesat four BorealEcosystem-Atmosphere Study(BOREAS) sites. Comparisonswere made betweenthe tower-basedbroadbandhemisphericalNDVI values and the narrowbandnadir-viewedNDVI valuesfrom helicoptermodular multiband radiometer(MMR). The comparisons indicatethat the tower NDVI valuesare closeto the MMR NDVIs in value for the BOREAS sites,but the range in tower NDVIs is not so great as in the MMR NDVIs. In 1996, BOREAS towersoperatedfrom before thaw to freeze-up,allowinga completepictureof growingseasonNDVI for fen, youngjack pine, black spruce,and aspensites.The tower-basedNDVI time seriesdisplaydifferent patterns for eachvegetationtype, showingthe effectsof snowcoverand vegetationgreen-upand senescence. Changesin solarzenith anglesare shownto have little effect on the seasonal NDVI patterns. 1.  Introduction  The seasonalchangein the reflectanceof landscapeshas long been an important area of studyin remote sensing.The spectralreflectanceof the landscapeis often expressedas an algebraiccombinationof spectralbandscalled vegetationindices.The most commonlyusedvegetationindex in seasonal studiesis the normalizeddifferencevegetationindex (NDVI). The pattern of seasonalchangeof NDVI providesan important descriptionof the vegetationand has been used to describea varietyof landscapefeatures,includingvegetationtype and productivity. Landsat data have been used to study the phenologyof NDVI. Multitemporal Landsatobservations were used to determine vegetationtype and productivityfor land coverssuch as crops [Hall and Badwar, 1987;Badwar, 1984] and forests [Blair and Baumgardener, 1977;Badwaret al., 1986]. With a 16-18 day period between observationsof a given location, Landsatis not able to generatefrequentobservations of vegetation seasonality.This weaknesshas been addressedin a studyof the boreal forestby usinggrowingdegreedaysinstead of time and combiningdata from multipleyears [Hall et al., •Universityof Maryland,CollegePark, NASA GoddardSpace  1991; Goetzand Prince,1996]. In the First International Satellite Land SurfaceClimatologyProject (ISLSCP) Field Experiment (FIFE) a detailed multitemporaltime seriesfor a grasslandwas created by combiningNDVI data from both Landsatand SPOT satellites[Hall et al., 1992a;Goetz,1997]. Few analysesof this type have been producedbecauseof the difficultiesand costof acquiringandprocessing the largenumber of Landsat and SPOT imagesneeded to get a detailed phenology. The studyof seasonalNDVI data blossomedwith the use of data from the advanced very high resolution radiometer (AVHRR) that flieson the NationalOceanicandAtmospheric Administration(NOAA) seriesof satellites.AVHRR observationsprovide views of the entire land surfaceof the Earth almost every day. Different vegetation types were found to have characteristicNDVI patterns through the year in the AVHRR data.Justiceet al. [1985]usedAVHRR data to study the phenologyof vegetation in South America, Africa, and southernAsia. Also, monthly NDVI values through the year were summedand comparedwith existingmaps of vegetation patterns.Tuckeret al. [1985] classifiedvegetationcovertypes for Africa. Goward et al. [1985] looked at North American vegetationpatterns.Both found a correspondence betweenthe integrated NDVI distributionsand the vegetation patterns. The  Flight Center, Greenbelt, Maryland.  2Department of SoilScience, University of BritishColumbia, Van-  seasonal  variations  in the AVHRR  NDVI  values  have  been usedto classifyvegetationglobally[DeFriesand Townshcouver, Canada. 3Institute of EcologyandResource Management, University of Ed- end,1994],for the United States[Lovelandet al., 1991],andfor the BOREAS studyarea [Steyaretet al., 1997]. inburgh,Edinburgh,Scotland. 4Department of Geography, Queen'sUniversity, Kingston, Ontario, Beyond estimatingcover type, it has been found that the Canada. NDVI valuesfrom AVHRR summedover a year were highly SNASAGoddardSpaceFlightCenter,Greenbelt, Maryland. correlatedwith net primary productivityfor major biomesin Copyright1999 by the American GeophysicalUnion. North and SouthAmerica [Gowardet al., 1985; Gowardand Dye, 1987;Gowardet al., 1987].The AVHRR NDVI data have Paper number 1999JD900164. 0148-0227/99/1999 JD900164509.00 also been used to determineinput parametersfor vegetation 27,935  27,936  HUEMMRICH  ET AL.: NDVI PHENOLOGY  productivitymodels[e.g.,Runningand Hunt, 1993;Princeand Goward, 1995;Field et al., 1995]. Other uses of seasonal variations  in NDVI  have shown cor-  relationsto atmosphericcarbondioxideconcentrations [Tucker et al., 1986]. Also, changesin the length of the growing seasonat high latitudeshave been detectedover an 11 year period in the AVHRR data. Thesechangesmay be related to vegetationresponseto global climate change[Myneniet al., 1997]. While the AVHRR NDVI data havebeenveryusefulfor the analysisof phenology,there are a number of difficultiesin usingthesedata. Somesourcesof error includecloudcontamination and atmosphericaerosols,gases,and water vapor. All of these error sources tend to cause a decrease in the value of  NDVI;  so to minimize their effect, the AVHRR  data are  compositedby selectingthe largest value of NDVI over a selectedtime period [Holben, 1986]. Generally,the data are compositedover periods of 10-30 days.This limits the temporal resolutionof the data and causesa bias toward higher valuesduringperiodsof change,suchas the start and end of the growingseason[Los et al., 1994;Holben, 1986]. The observationschosenin the compositing processmayhavevarying view and solaranglesfrom pixel to pixel, and thesevariations can effect the NDVI value [Gowardet al., 1991;Gowardand Huemmrich,1992]. AVHRR data are also difficult to comparewith surface observations due to the largepixel size.The observations have, at best, a 1 km resolution. However, because of variations in  pixel size from different view anglesand uncertaintiesin locatingthe multiple pixelsin a temporal seriesthe actualresolution may be 3-4 km [Steyaretet al., 1997]. The limitations on both temporal and spatial resolutionof satelliteobservations of phenologymakeit difficultto compare directlywith surfaceobservations of plant processes. There are alsofew studiesthat have made frequent surfaceobservations of vegetationreflectanceto be able to produce a detailed ground-based NDVI phenology.One of the reasonsdetailed ground-basedNDVI phenologieshave not been collectedis that most spectroradiometers used in remote sensingfield work are not designedto operate automaticallyand to be exposedto the weather.Thus theseobservations require personnel to go out and make the measurements,limiting the number  of observations  that can be collected.  In the case of  forests there is also the problem of getting the instruments abovethe canopy. As part of surface flux studiesof forests,instrumentsto measure  the radiative  flux have been mounted  on towers to  make continuousobservationsof the forest canopy.AVHRR NDVI  values have been shown to be correlated with albedos of  photosynthetically activeradiation(PAR) measuredthroughout a year abovea red oak forestin Massachusetts [Sakaiet al., 1997]. PAR and shortwavealbedos have been shown to be sensitiveto seasonalchangesin leaf area indexfor this forest [Mooreet al., 1996].Sakaiet al. [1997]alsoshowthat the PAR albedois relatedto the canopyresistanceto water vapor transport, linking radiativemeasurementsand surfacefluxes. The Boreal Ecosystem-Atmosphere Study (BOREAS) experiment plan establishedtowers in severaldifferent boreal vegetationtypesto collect surfaceflux and meteorologymeasurements[Sellerset al., 1995, 1997]. BOREAS provided the opportunityto producedetailedNDVI phenologiesof several boreal vegetationtypesby usingexistingmicrometeorological instrumentsin a novelway. The instrumentationon severalof  thesetowersincludedupward and downwardviewingsensors for PAR  and shortwave  solar radiation.  These common  mi-  crometeorologicalradiation sensorswere used to calculate broadband  NDVI  values. As these sensors collected data con-  tinuously, it was straightforwardto produce surface NDVI phenologiesfor specificvegetationtypeswith a dailytime step.  2.  Methods Data from four BOREAS  flux tower sites were used in this  study.These sitesrepresentfour very different boreal vegetation cover types:mature aspen,mature black spruce,young jack pine, and fen. The old aspenand old black sprucesites were in the southern study area in Saskatchewan,and the youngjack pine and fen siteswere in the northernstudyarea in Manitoba.  Each site had a tower that extended above the  canopyfor supportingeddy correlation and meteorological instrumentation.All of thesesiteshad upwardand downward viewing sensorsmeasuring PAR and shortwaveradiation mountedabovethe canopy.All towersusedLi-Cor quantum sensorsthat measurethe numberof photonswith wavelengths between0.4 and 0.7 t•m. The aspen,youngjack pine, and fen sitesused Eppley pyranometersthat measurethe shortwave solar irradiance between 0.285 and 2.8 t•m. The old black sprucesite used a Kipp and Zonen pyranometerthat has a similar wavelength response as the Eppley pyranometers [Jarviset al., 1997;McCaugheyet al., 1997;Lafieur et al., 1997; Black et al., 1996]. The old aspensitehad a matureaspen(Populustremuloides) overstorywith a hazelnut (Coryluscornuta)understory.The aspentreeswere about 60 yearsold, -20 rn in height,with a midsummercanopy cover of-90%. The instrumentswere positionedat 36 m abovethe ground[Blacket al., 1996]. The old black sprucesite was dominatedby mature black spruce(Piceamariana) with some tamarack(Larix laricina) occurringat the site.The sprucetreeswere over 100yearsold. The understoryconsistedmostlyof feather moss(Pleurozium schreberi)and Labrador tea (Ledurn groenlandicum).Tree height was around 10 m, with canopycover of -55%. The instrumentswere mountedat 16 rn abovethe ground[Jarviset al., 1997]. The youngjack pine siteconsistedof jack pine (Pinusbanksiana),whichwere lessthan 25 yearsold and onlyabout2.5 m tall. The understorywas coveredmostlyby lichens(Cladina spp.). Instrumentswere attached to the tower at -11.5 m aboveground[McCaugheyet al., 1997]. The fen sitewas a wetlandwith no vegetationtaller than a meter.Groundcoverincludedbogbirch(Betulapumila),buckbean (Menyanthes trifoliata),and sedges(Corexrostrata).Instrumentswere mounted--•10.3rn abovethe ground[Lafieuret al., 1997]. Frequent observationsof NDVI have been derived using data from the radiation  sensors mounted  on the flux towers.  The normalizeddifferencevegetationindex(NDVI) isdefinedas  NDVI = (PNIR-- PV,s)/(PNIR q- PV,s),  (])  where Pvisis the reflectancein the visiblewavelengths,and PNIRis the reflectancein the near-infraredwavelengths. In this study,measurementsof incident and reflected PAR and incident and reflected  shortwave radiation  were used in the calcu-  lation of NDVI. First, the PAR measurementswere converted  from micromolesof photonsto joulesby multiplyingby 0.25 J  HUEMMRICH  ET AL.: NDVI PHENOLOGY  /xmol-•.Thisconversion factorisbasedon theenergyof pho-  27,937 BOREAS  Helo  MMR  NDVl  tonsof greenlight and convertsthe PAR data to units of watts  m-2. PAR reflectance (PPAR)wasthe ratio of reflectedand  y=0.23996 +0.69734x  incomingPAR (EpARrefland EpARin respectively), PPAR '-- EpARrefi/EpARin ß  a "øø"'"-  09 R^2 =0.904  (2)  O8  To calculatean opticalinfraredreflectance(POIR),the PAR  •  irradiance values were subtracted from the shortwave irradi-  •  0.7  ances forboth incoming and reflected fluxes. Optical infrared• o.6  reflectance wasthe reflected dividedby theincoming of the  •,  o.s  differences oftheirradiances. From theradiation instruments,  PAR and optical infrared reflectanceswere calculated,  1OOiR •- (mswre fl - mpARrefl)/(mswinmpARin)  (3)  where Eswrefl is the shortwaveradiant exitanceand Eswin is the shortwave  irradiance.  These reflectances  NDVI = (JOOi R -- jOPAR)/(JOOi R q- jOPAR).  0.2  0.• (4)  NDVI values calculated from Landsat thematic mapper (TM) data useTM channels3 and 4 with wavelengthbandsof 0.622-0.699 /xm and 0.771-0.905 /xm, respectively.AVHRR NDVI values use AVHRR channels1 and 2 with spectral rangesof 0.58-0.68 /xm and 0.725-1.1 /xm. The broadband NDVI calculatedwith the micrometeorologicalsensorsusesa PAR wavelengthband of 0.4-0.7/xm and an opticalinfrared band that is effectivelybetween 0.7 and 2.8 /xm along with a 0.28-0.4 /xm addition. Becausemost of the radiant energy, both incident and reflected,in the optical infrared band is in the lower wavelengthsof that band, i.e., in a similar spectral range asthe AVHRR channel2, we can expectthe broadband NDVI to respondto the samevegetationeffectsas the satelNDVIs.  To evaluate  0.3  were substituted  into(1) to calculate a broadband NDVI,  lite-derived  Z  0,0  0.0  0.1  between  the broadband  0.3  0.4  0.S  NDVl  from  Narrow  0.6  0.7  0.8  0.9  10  Bands  Figure 1. BroadbandNDVI valuescomparedto narrowband NDVI values. Data from the helicopter-mountedMMR for over 60 BOREAS sitescollectedthrough the growingseason of 1994. The  broadband  NDVI  used combinations  of MMR  bands 1, 2, and 3 for the visible band and MMR bands 4-7 for the near-infrared  band in the NDVI  calculation.  The narrow-  band NDVI was calculatedusing only MMR bands 3 and 4. The dashed line is the 1-to-1 line; the solid line is the least  squaresfit to the data.  broadbandreflectances,and from the reflectances,NDVIbb was determined.  the differences  0.2  The  broadband  MMR  reflectances  do not  NDVI  cover the full wavelengthrange of the micrometeorological and a typical narrowbandNDVI, an analysiswas performed instruments;however,most of the gapsbetween the MMR usingthe helicopter-mountedmultibandmodular radiometer near-infraredbandsoccurin spectralregionswhere absorption (MMR) data collectedoverthe BOREAS sites[Loechelet al., by atmosphericwater vapor allowsvery little energyto reach 1997]. The helicopter made nadir observationsof over 60 the Earth. BOREAS sites, representingthe different vegetationcover Figure 1 showsthat there is a high correlation between typesfound in the studyarea. These data were collectedat broadbandand narrowbandNDVIs, with an R 2 of 0.9. The solar zenith angles ranging from 32ø to 62ø throughout the slope of the regressionline indicates that the broadband growingseasonof 1994.MMR datawere reportedasreflected NDVIbb showsa smallerrange in valuesthan the narrowband radiance and a reflectance for each channel; from these data  NDVInb. The highcorrelationbetweenthe two differentNDVIs indicatesthat the widening of the bands does not introduce significantamountsof noisefor the BOREAS sites,althoughit ues were then calculated. The first was the narrowband NDVI doesdecreasethe range of responseobserved. (NDVInb) whichusedreflectances from MMR channels3 and NDVI values from the micrometeorologicalinstruments 4. To calculatea broadbandNDVI (NDVIbb), the radiances were calculatedfor the times of helicopterobservationsof the for channels 1-3 and channels 4-7 were summed. The incomsites. Figure 2 compares the MMR and flux tower NDVI ing and reflected multiband radianceswere used to calculate  the incomingirradianceswere determined(see Table 1 for bandwidthsof each MMR channel).Two different NDVI val-  Table  1.  Barnes  Multiband  Module  Radiometer  Bandwidths  (NDVIrad). This figure showsthat the MMR narrowband NDVInb appearsto be better correlatedto the MMR broadband NDVIbb than the flux tower broadbandNDVIrad. This is not too surprisingsincethe two different MMR calculations both come from the same data. Differences  Bandwidth, Channel Channel Channel Channel Channel Channel Channel  1 2 3 4 5 6 7  0.45-0.52 0.51-0.60 0.63-0.68 0.75-0.88 1.17-1.33 1.57-1.80 2.08-2.37  in the NDVI  val-  ues are due to different areasbeing viewed by the MMR and the flux tower instruments  and that the MMR  is a bidirectional  reflectancemeasure,while the flux tower data representhemisphericreflectances.The MMR givesa slightlyhigherbroadband NDVIbb than the flux tower NDVIra d. The helicopterMMR data showthat NDVIs calculatedfrom broad spectralbands respond to variationsin vegetation in much the sameway as narrowbandNDVIs.  27,938  HUEMMRICH Narrow  and  Broad  BOREAS  Band  ET AL.: NDVI PHENOLOGY  observed incident PARwaswithin40 W m-2 of thepredicted  NDVI  value were consideredclear days.This method must be used with care near the end of the growingseasonwhen the incident PAR valuesbecomefairly small and the differencesbetween clear and cloudydaysshrink.In Figures3-6 the NDVI points for every day are plotted with dashedlines, with solid lines connectingthe pointsfor clear days.After removingcloudy days,approximately a third of the pointsremain.Thesepoints producea smoothseasonalNDVI curve. Figure3 showsdailyNDVIs from 1996for the old aspensite. To have shortwaveirradiancedata for this period, data from two differentpyranometers were combined.Both instruments  TF Sites  0.8  0,7  0.6  •,,  z  0.5  •  0.4  m 0.3 0.2  / 0.1  0.0  I" MMRBroad Band NDVi J ß TFBroad Band NDV  X  were mounted on the same tower, and when both were in  ßTF Broad Band NDV i  0.0 ß011' 012' 013' 014' 0'.5' 0'.6' 017' 0.8 MMR  Narrow  Band  operation,their measurements comparedwell with eachother. The negativevalues of NDVI at the beginningof the time series were due to snow cover. The snow thawed and the NDVI  NDVl  increasedto about0.35 at day 134 (May 14). Thiswasa period Figure 2. The x axisis the narrowbandNDVI from the he- of baregroundandleaflesstreesthat lasteduntil day145(May licopter-mounted MMR, asdescribedin Figure1. The squares 25) when the NDVI curvebegana suddenincreasewith the are the corresponding broad band NDVI valuesfrom the startof the growingseason.NDVI increasedover the following MMR data. The diamonds are the NDVI values from the micrometeorological instruments,calculatedfor the times of 20 dayperiodup to 0.71.The NDVI continuedto increasebut helicopterobservations of the sites.The line is the 1-to-1line. at a muchslowerpace,to a maximumvalue of 0.78 at day219  3.  Results  NDVI phenologies were producedfor the sitesat a daily time step.The datawere reportedashalf hourlyaverages,and the valuesreportedfor 1400LT for eachdaywere used.This time waschosento approximatethe time of a NOAA satellite overpass.Data collectedon cloudydayscan producespikes, generallyshowinglarger NDVI values,in the time series.To eliminatethesespikes,time seriesbasedonlyon datacollected on clear dayswere created.Clear dayswere detectedby predictingincidentPAR for clear daysbasedon the solarzenith angle at the time of the observations; then any day where the  (August7). It thenbegana slowdeclineto day269(September 26) overwhichit droppedto 0.70.After thisdaytherewasan accelerating declinewith a suddendrop-offafter day277 (October 1). The earlypart of the declinein NDVI seemeddueto an increasein PAR reflectancewhich may be due to leaf color changes,and the suddendrop-off may occurwith the lossof leaves causingdramatic changesin both the PAR and the infraredbands.After the autumnNDVI drop,it settledaround a valueof about0.39for days286-297 (October13-24). This is anotherperiodof leaflesstrees,but this autumnperiodhad a higherNDVI than the springtimeleaf-off period.This may be becausethe newlyfallen leaveshave not broken downyet and are still scatteringnear IR radiation causinghigher ND-  1.0  ?...':': :•?:..,.,L..'. .... .3.,.:  .  0.5  l:  ß:  l  0.0  .................... 1996 All Points  ß  -0.5 i 90  1996 Clear Sky  ,,  1  20 ' ' 150i , ß180i ß I 210J  2;0  ß  '  i  ,  270  i  i  ,  300  I  I  '  330  I  36O  Day of Year  Figure 3. Daily NDVI observations of the BOREAS southernstudyarea old aspensite for 1996. Observationsfor 1400 LT. Dotted line connectsall observations.Solid line connectsclear-dayobservations,plotted as solid diamonds.  HUEMMRICH  ET AL.: NDVI PHENOLOGY  27,939  1.0  ..  .  !: .. .....,  • •..",.  _•  0.5  0.0  ' I  "  ;  .......... 1995 AllPoints  •  •  • /, -0.5  '.!  -- 1995 Clear Sky .......... 1996 AllPoints  ,  90  1994 Clear Sky  •  ,  120  ,  150  1996ClearSky  ,  180  .  .  210  ,  .  .  ,  240  270  .  .  E  .  300  .  330  Day of Year  Figure 4. Daily NDVI observations of the BOREAS northernstudyarea fen sitefor 1994,1995,and 1996. Observationsfor 1400 LT. Dashedlines connectall daily observations. Solid lines connectclear-dayobservations,plotted with symbols.  VIs. After that, aroundday300 (October26) there is another suddendrop whichmay be due to snowfall. This seasonalNDVI pattern is similarto the one observed with helicopterdata of aspenforestsin the SuperiorNational Forest in Minnesota[Hall et al., 1992b].However,the NDVI phenology, basedon monthlycomposited AVHRR datashown by Steyaertet al. [1997],doesnot showthe detailedfeatures  seenin the flux tower NDVI curve.The AVHRR data simply showa steadyincreasein NDVI from April to June, missing the stepthat occursafter snowmeltand before green-up. The other deciduoussite observedin this study is the fen, and it has a different seasonalNDVI pattern than the old  aspen site.Figure4 shows theNDVIsforthefenfor 1994, 1995, and 1996. In 1995 a limited amount of data was collected,  NSA-YJP 1994  NDVl &  1996  1.0  i  .................... 1994All Points  i  ß  i[  0.8  1994ClearSky  .......... 1996 All Points • 1996ClearSky  ,: ,,  0.6  0.4  0.2  0.0  -0.2  i  20  ß  150  .  i  180  .  .  i  210  !  240  i  270  i  300  330  Day of Year  Figure 5. Daily NDVI observations of the BOREAS northernstudyareayoungjackpinesitefor 1994and 1996. Observations for 1400 LT. Dashedlines connectall daily observations. Solid lines connectclear-day observations, plotted with symbols.  27,940  HUEMMRICH  ET AL.: NDVI  PHENOLOGY  1.0  t  0.5  0.0  .................... 1004 All Points  *  -0.5 60 . ß90,  1994 Clear Sky  ..........  1996 All Points  •  1996 Clear Sky  • . . 150, ß . 180, ß ß210, ß . 240, ß ß270, ß ß300, ß . 330, . . 360  1 0  Day of Year  Figure 6. DailyNDVI observations of theBOREASsouthernstudyareaoldblacksprucesitefor 1994and 1996.Observations for 1400LT. Dashedlinesconnectall dailyobservations. Solidlinesconnectclear-day observations, plotted with symbols.  capturingonlythe thawperiod.Note that the NDVI valuesfor 1994 and 1996 are very closein value throughthe growing season.This is an indication,alongwith the generalsmoothnessof the curvesfrom the clearskydata,of the repeatability of this methodology. As with the aspensite,all of the NDVI curvesbeganwith very low values, around -0.4, indicative of snow cover. With snowmelt,NDVI rosequickly;in 1996the snowmeltlaggedthe other two yearsby over 10 days.With the thaw of the snow,  perioddueto briefspringtime snowfalls. Aroundday142(May 22) for both yearsthere was a small dip, followedby the beginningof green-up.In 1996,NDVI increasedfrom 0.46 to a midsummerhighof 0.73 over44 days.The NDVI remained fairly constantuntil day 231 (August19) then beganslowly decreasing out to day283 (October10). Followingthat,NDVI decreased from 0.57 to 0.04 due to snow.  The seasonal patternfor the fen is differentthanthe aspen (see Figure 7). In the fen, green-uptakes longer,44 days NDVI reached a value of around 0.5. In 1994 and 1996 there comparedto the aspen's20 daygreen-up.The fen beginsthe weresharpdownwardspikesin the unfiltereddataduringthis green-upwith a higherNDVI, 0.46, than the leaflessaspen  1.0  0.5  0.0  *  Aspen  --  Jack Pine  •  •  60  90  120  150  180 2;0  Fen  Black Spruce  240  270  300  330  360  Day of Year (1996)  Figure 7. NDVI observations for clear daysof four BOREAS sitesfor 1996. Observationsfor 1400 LT.  HUEMMRICH  ET AL.: NDVI  NDVI of 0.35. During the springtime green-up, both sites reach a maximumNDVI of about 0.73, but the aspencontinues a slowincreaseup to 0.79, while the fen remainsat that value. The pattern of the NDVI changeduring the green-upof the fen is similarto that of a grassland,asseenin FIFE [Hall et al., 1992; Goetz, 1997]. Turning from the deciduousfen site to the evergreenyoung jack pine site,there is againa differentNDV! phenology(Fig-  ure 5). Noticethatthe late-spring snowfallon day141(May 21) of 1996 showedup as a downwardspike, and a similar spike occurredin the data for the nearbyfen site (Figure 4). Also, both siteshad a largedrop-offfollowingday285 (October12). Both of theseeffectswere mostlikely due to the two sitesbeing hit by the same winter storms.Following the snowmelt,the NDVI curvefor the youngjack pine sitewasbasicallyflat. The jack pines went through a seasonalchange,but it was very subtle. In 1996 there was a green-up period from days 143 (May 23) to 223 (August11), where NDVI went from 0.56 to 0.63. After that the NDVI decreasedvery slightlyuntil day 286 (October 13) when a snowfallcauseda large drop. Although another site dominatedby conifers,the old black sprucephenologydiffersfrom the youngjack pine (Figures6 and 7). The blacksprucesitehad a largerseasonalvariationin NDVI andconsistently smallerNDVI valuesthan thejack pine site. The old black sprucesite began springtimegreen-up on day 113 (April 23). This initial increasemay havebeen due to snowmelt.At that time the NDVI value was 0.24, higher than the other siteswhensnowcovered.The highNDVI maybe due to the instrumentsviewing a mixture of the green trees and snow-covered ground.From day 113 there was an increasein NDVI up through day 180 (June 29), when NDVI reached 0.55. The curveremainedflat for about40 daysand thenbegan a slowdecreaseout to day 275 (October2) when it startedto drop off faster.NDVI remainednear 0.36 from days281 (October 8) to 313 (November9). Followingthat period, NDVI valuesdropped,probablydue to snowfall.There was a sudden upward jump between days 318 and 321 (November 14 and 17), andfollowingthat rise,NDVI valuesreachedtheir highest valuesof the year. This may be due to the meltingof the snow, coupledwith the effect of large solarzenith angles.After day 328 (November24) the NDVI valuesbecamenegative.These values were very low becausethe instrumentswere viewing snowin the trees as well as on the ground. The youngjack pine data suggestthat the seasonalchangein NDVI for conifersis very small.If that is true for spruce,what accountsfor the changesat the old black sprucesite? Measurementsof the spectral reflectanceof understorycomponentsat this sitewere made in 1994 [Miller et al., 1997].These measurementsshowa changein NDVI of the understoryfrom springto summerthat is approximatelyas large as seenin the flux tower NDVIs. This observedchange in the background reflectancemay accountfor someof the seasonalvariation in NDVI  for this site.  The NDVI values,generatedusingmicrometeorologicalinstruments,show consistencyin day-to-dayclear sky observations,aswell as in data collectedin differentyears.Within the data, similaritiesbetweensiteswere observedin the response to snow-coveredground. These resultssuggestthat this approach producesa measurementof NDVI that is consistent over time andbetweensites.The NDVI phenologiesproduced with this methodprovideboth a spatialand a temporal detail unavailable  from satellite  data. The BOREAS  data show that  for four different boreal land covertypes,eachhas a different  PHENOLOGY  27,941  seasonalNDVI pattern in both shapeand midsummermaximum NDV! (Figure 7). 4.  Discussion  The detailedNDVI phenologlespresentedhere providethe opportunity to considersome of the causesof the seasonal patterns. Two examplesare consideredhere, one looking at rates of change in leaf area in deciduoussites and the other examiningthe effectsof varyingsolar zenith angles. In the deciduoussitesthe period of most dramaticchangeis during the springgreen-up.In the boreal forest the transition from winter senescence to springtimegrowthis a critical time. During this period the aspenand fen sitesdisplayeddifferent patternsof NDVI over time. The differencesbetweenthe sites are clearlyvisiblein the flux tower data, althoughthe details would be smearedout in AVHRR phenologies.The aspenand fen NDVI patternsduring green-upsuggestdifferent growth strategiesfor the vegetation of these two sites, one forested and one predominatelyherbaceous. One method for exploringthe relationshipbetweenNDVI and vegetationcharacteristicsis through the use of radiative transfer models.Using the SAIL (scatteringfrom arbitrarily inclined leaves) radiative transfer model the variations in NDVI with leaf area index (LAI) can be calculated[Verhoef, 1984; Gowardand Huemmrich,1992]. Through the model the effectson NDVI of three different modesof LAI growthwith time havebeenconsidered: linear,logarithmic,and exponential. Changesin LAI over an arbitrarytime periodfor eachof these modelsof leaf growthare shownin Figure8a-8c, alongwith the resultingNDVI valuesfrom the SAIL model.In thesesimulations all factorssuchas leaf and backgroundopticalpropertiesand canopystructurewere held constant,and only LAI wasvaried. This isolatesonly the effectsof differing rates of leaf growth. If the leaf area increaseslinearly with time, NDVI increases very rapidly at first, then flattens out as it beginsto reach a saturationvalue (Figure 8a). A logarithmicincreasein LAI, again producesa very rapid initial increasein NDVI, which then flattensout to a saturationvalue (Figure 8b). For exponential LAI growth,NDVI followsa more S-shapedtrajectory, with the maximum NDVI valuesbeing reached more toward the end of the time period (Figure 8c). This exerciseindicates that when examiningNDVI data, it would be difficult to determine if the growthrate of leavesis linear or logarithmic,as the basicshapeof the NDVI curvesare very similar for these two  cases.  Comparing the SAIL model resultswith the NDVI observationssuggeststhat the springtimeleaf growth of the aspen standwas linear or logarithmic,while the herbaceousfen site had more of an exponentialgrowthpattern. Further observationsof green-upin other yearsexperiencingdifferentweather conditionsor other siteswith different vegetationtypescould further illuminate the differencesin the growth rate of leaves in the spring. Seasonalvariationin NDVI maybe due to factorsother than changesin the vegetation,suchassolarzenith anglevariations. In general, the seasonalchangesin solar zenith angle are similarto the seasonalcourseof deciduousvegetationgrowth, making it difficult to separatethe effectsof the two factors. Over the period of the NDVI data presentedin this paper the solar zenith angle at 1400 LT ranged between 33ø and 79ø. Leblancet al. [1997]modeledthe effectof solarzenith angleon nadir-viewed  NDVI  for BOREAS  sites. Their  results show a  27,942  HUEMMRICH  4.0  ET AL.: NDVI PHENOLOGY  ,7  •.6 0.8  0.6  Moore et al. [1996] performed an analysisof the effect of  solar zenith angle forshortwave albedo ofared oak forest by  .  examining data collected over individual days selected throughout the year. They found that shortwavealbedo in-  -s  creased slightly withsolarzenithanglefor solarzenithangles  .  lessthan 70ø.They alsoobserveddifferencesin albedobetween  .  morning andafternoon observations withthesamesolarzenith  i • angles.  Usingthe sameapproachasMooreet al. [1996],the effectof  0.4]/ /•  i'2 solar zenith angle variation was examined for the black spruc site. As the tower flux radiation  sensors collected  data contin-  uously,this allows NDVI valuesto be calculatedthroughout the day. Thus NDVI values can be determined over a wide  0.e  0.0  •0 range of solar zenith angles within asingle day. We assu  1.0 t ]/ •  i solstice, maximizing the range of solar zenith angles. The PAR iszero. NDVivalueswereve  that the variationsin NDVI within the singleday are not due to changesin the vegetation.Figure 9 showsthe half hourly valuesof NDVI and incidentPAR for day 165 (June 14). Day 165 was chosenbecauseit was a clear day near the summer data providean indicationof the courseof the Sunthroughthe  0.8  day,withsunrise andsunset occurring whentheincident PAR  s  ø'eli •/• V ....  0.2  clear day values for all of 1996 in Figure 10. There are small  differencesbetween the morning and the afternoon NDVI values at the lower solar zenith values. The differences  0.21 /  0.0  theafternoon. TheNDVIvalues areplotted against solar i •, into general downward trend in the NDVI through the midda and zenith anglefor the half hourly data from day 165 and for the  0-4t/ 0.o  !'••  0.4  NDVI I  ] ;. cai I b o1  OiS  018  •  •.0  to illumination differences caused by variations inthe azimuth angle between morning and afternoon. Beyond asolar solar  zenithangleof 70ø thefluxtowerNDVI begins to increase rapidly.Figure 9 displaysmuchlessof an effect of solarzenith angle on the tower flux NDVI comparedto the resultsof Leblanc et al. [1997]. The disagreementmay be due to the  NDVI[  ,;  widen  untilat70øthedifference isnearly0.1NDVI. Thismaybedue  lO  0.8  fairlyconsistent through themiddle oftheday.There wasa  LAII  6  differences  between  a nadir view NDVI  and the tower  flux  hemisphericalNDVI. The NDVI variationswithin a daycanbe contrastedwith the 0.6 seasonalvariations to separate the effects of varying solar zenith anglefrom changesin the landscape.Throughmostof the year the daily values fall in the range of values from day 0.4 165. In the springand autumn,valuesfall away from the half hourlyvalues,suggesting changesin the landscaperather than effect of illumination variations. The largepeak that occurred 0.2 followingday 318 is shownin Figure 10 to be within the range of valuesfrom day 165, indicatingthe large NDVI valueswere due to the large solar zenith anglesat that time of the year. 0.0 • 3 I T C o O0 0.2 0.4 0.6 0.8 1.o Combininghalf hourly and daily valuescan be usefulin separating out illuminationeffectsfrom changesin the landscape, Time improvingthe usefulnessof thesedata in detectingbiological Figure 8. (a) Leaf area index,plotted as diamonds,varying changes. linearlywith time, and the resultingvariationin NDVI, plotted The small effect of variation in solarzenith angle on NDVI as squares;(b) LAI varyinglogarithmicallywith time and the for a wide rangeof solarzenithanglesat thissitesuggests that if resultingvariation in NDVI; (c) LAI varying exponentially needed,more dailyvaluescouldbe addedto the seasonalcurve. with time and the resultingvariation in NDVI. All NDVI Valuescouldbe filled in for cloudydaysthat had beenremoved values calculated from the SAIL model. by using data collected during clear periods at other times duringthe day,aslongasthe solarzenithanglewasnot too large. monotonicincreasein NDVI with solarzenith angle.Over the range of solarzenith anglesobservedin thesedata the model 5. Conclusions indicateda changeof 0.18 in NDVI. This raisesquestionsin A method was describedwhich can determine NDVI using interpretingthe NDVI data;however,the useof the radiation sensorsprovidesa method to examinethis effect. existingmicrometeorologicalinstruments.This techniqueal,  HUEMMRICH  ET AL.: NDVI  PHENOLOGY  27,943  5OO  0.8  a• Incident PAR 4OO 0.6  0.4  0.2  0.0  8  10  12  14  16  18  GMT  20  22  24  26  28  Time  Day 165  Figure 9. Half hourlyvaluesof NDVI, plottedasdiamonds, andincidentPAR, plottedassquares, for day 165(June14, 1996)for the BOREASsouthernstudyareaold blacksprucesite. lowed the constructionof temporallydetailedNDVI phenologiesfor specificlocations.NDVI phenologies generatedfor four different sites within the boreal forest biome showed that  each different land cover has a unique form. 1  0.5  This techniquecan help to improveunderstandingof the timingand natureof seasonaland multiyearvariationsin vegetation. Gouldenet al. [1996] note that total seasonalcarbon uptakeof deciduous forestsstronglydependson the timingof leaf out. Attemptsto useremotelysensedapproaches to accurately determineannual carbonbudgetsrequire an understandingof seasonalphenologies with a highertemporalresolutionthan canbe presentlyprovidedby satellitedata.NDVI from micrometeorological instrumentscan beginto developa better understandingof the timing of seasonalvegetation changes.The use of NDVI providesa direct link between surfaceobservations and satellitevegetationindex data. Acknowledgments.The authorswish to thank all who were involvedin the collectionand processing of the data usedin this study. The instrumentsat the SSA OBS sitewere installedand monitoredby S. E. Hale andS. L. Scottin 1994andby M. B. RaymentandS. L. Scott in 1996.The datafrom bothyearswere qualitycontrolledand submitted to the databaseby J. M. Massheder.Instrumentinstallation,data collection,and data processing for the NSA YJP and fen siteswere performedbyP. Lafleur,D. Joiner,P. Bartlett,A. Costello,B. Mantha, K. Boudreau,B. Robertson,D. Mueller, and L. Liblik. SSA OA data  '.o  were collectedwith the assistanceof Z. Nesic, U. Gramann, R. Ketler, -0.5  .  30  •  40  '  •  '  50  Solar  •  60  Zenith  '  •  '  70  P. Pacholek,and A. Barr. Thanks also go to the Saskatchewan Research Council, John Miller, and the RSS-19 team and Charlie  •  80  90  Walthall  and the RSS-3 team for the use of their data.  Angle  References Day 165 ........ O........ 1996  Badhwar,G. D., Use of Landsat-derivedprofile featuresfor spring small-grains classification, Int. J. RemoteSens.,5(5), 783-797,1984.  Figure 10. NDVI versussolarzenith angleat the BOREAS Badhwar, G. D., R. B. MacDonald, F. G. Hall, and J. G. Cames, southernstudyarea old blacksprucesite.Half hourlyvaluesof Spectralcharacterization of biophysical characteristics in a boreal NDVI from day 165 (June 14, 1996), plotted as circlesconforest:RelationshipbetweenThematicMapper band reflectance nectedby a solidline, and 1996dailyclear-dayNDVI observa- and leaf area index for aspen,IEEE Trans.Geosci.RemoteSens., 24(3), 322-326, 1986. tionsfor 1400LT, plottedasdiamonds connected bya dottedline.  27,944  HUEMMRICH  ET AL.: NDVI PHENOLOGY  Black,T. A., et al., Annual cyclesof watervapourand carbondioxide fluxesin and abovea borealaspenforest,GlobalChangeBiol.,2(3), 219-229, 1996.  forests:A model interpretationof measurements, Can J. Remote Sens.,23(4), 369-380, 1997. Loechel, S. E., C. L. Walthall, E. B. de Colstoun, J. Chen, B. L.  Markham, and J. Miller, Variability of borealforestreflectancesas Blair, B. O., and M. F. Baumgardner,Detectionof green and brown measuredfrom a helicopterplatform,J. Geoœhys. Res.,102, 29,495wave in hardwoodcanopycoversusingmulti-date, multispectral 29,504, 1997. data from Landsat-l, Agron.J., 69, 808-811, 1977. DeFries, R. S., and J. R. G. Townshend,NDVI-derived land cover Los,S. O., C. O. Justice,and C. J. Tucker,A 1ø x 1øglobalNDVI data set for climate studies derived from the GIMMS continental NDVI classifications at a globalscale,Int. J. RemoteSens.,15(17), 35673586, 1994. Field, C. B., J. T. Randerson, and C. M. Maimstrom, Global net  data, Int. J. Remote Sens., 15, 3493-3518, 1994. Loveland, T. R., J. W. Merchant, D. O. Ohlen, and F. J. Brown,  Developmentof a land covercharacteristics databasefor the conprimaryproduction--Combiningecologyand remote sensing,ReterminousU.S., Photogramm. Eng.RemoteSens.,57(11), 1453-1463, moteSens.Environ.,51(1), 74-88, 1995. 1991. Goetz, S. J., Multi-sensoranalysisof NDVI, surfacetemperature,and biophysical variablesat a mixedgrasslandsite,Int. J. RemoteSens., McCaughey,J. H., P. Lafleur, D. Joiner, D. Jelinski,and M. Ryan, Magnitudesand seasonalpatterns of energy,water, and carbon 18(1), 71-94, 1997. exchanges at a borealyoungjack pine forestin the BOREAS northGoetz, S. J., and S. D. Prince,Remote sensingof net primaryproducern studyarea,J. Geoœhys. Res.,102, 28,997-29,008,1997. tion in boreal forest stands,Agric. For. Meteorol.,78(3), 149-179, 1996. Miller, J. R., et al., Seasonalchangein understoryreflectanceof boreal forestsand influenceon canopyvegetationindices,J. Geoœhys. Res., Goward,S. N., and D. G. Dye, EvaluatingNorth Americannet primary 102, 29,475-29,482, 1997. productivitywith satelliteobservations, Adv. SpaceRes.,7(11), 165174, 1987. Moore, K. E., D. R. Fitzjarrald, R. K. Sakai, M. L. Goulden, J. W. Goward, S. N., and K. F. Huemmrich, Vegetation canopyPAR abMunger, and S.C. Wofsy, Seasonalvariationin radiative and turbulent exchangeat a deciduousforest in central Massachusetts, sorptanceand the normalizeddifferencevegetationindex:An asJ. Appl. Meteorol.,35, 122-134, 1996. sessment usingthe SAIL model,RemoteSens.Environ.,39, 119-140, 1992. Myneni, R. B., C. D. Keeling, C. J. Tucker, G. Asrar, and R. R. Nemani,Increasedplant growthin the northernhighlatitudesfrom Goward,S. N., C. J. Tucker, and D. G. Dye, North Americanvegeta1981to 1991,Nature,386(6626),698-702, 1997. tion patternsobservedwith the NOAA-7 advancedvery high resolution radiometer,Vegetatio,64, 3-14, 1985. Prince,S. D., and S. N. Goward,Globalprimaryproduction:A remote Goward, S. N., A. Kerber, D. G. Dye, and V. Kalb, Comparisonof sensingapproach,J. Biogeogr., 22(4-5), 815-835, 1995. North and South American biomes from AVHRR observations, Running,S. W., and E. R. Hunt, Generalizationof a forestecosystem processmodelfor other biomes,BIOME-BGC, and an application Geocarto,2(1), 27-40, 1987. for global scalemodels,in ScalingPhysiological Processes: Leaf to Goward, S. N., B. Markham, D. G. Dye, W. Dulaney, and J. Yang, Globe, edited by J. R. Ehleringer and C. B. Field, pp. 141-158, NormalizedDifference Vegetation Index measurementsfrom the Academic,San Diego, Calif., 1993. AdvancedVery High ResolutionRadiometer,RemoteSens.Environ., 35, 257-277, 1991. Sakai,R. K., D. R. Fitzjarrald, and K. E. Moore, Detectingleaf area and surfaceresistanceduring transitionseasons, Agric.For. MeteoGoulden, M. L., J. W. Munger, S.-M. Fan, B.C. Daube, and S.C. rol., 84, 273-284, 1997. Wofsy,Exchangeof carbondioxideby a deciduousforest:Response to interannualclimatevariability,Science,271(52556), 1576-1578, Sellers, P. J., et al., The Boreal Ecosystem-AtmosphereStudy 1996. (BOREAS):An overviewandearlyresultsfromthe 1994fieldyear, Bull. Am. Meteorol. Soc., 76, 1549-1577, 1995. Hall, F. G., and G. D. Badwar, Signature-extendibletechnology: Global space-based crop recognition,IEEE Trans.Geosci.Remote Sellers,P. J., et al., BOREAS in 1997:Experimentoverview,scientific Sens.,GE-25, 93-103, 1987. resultsand future directions,J. Geoœhys. Res.,102, 28,731-28,770, Hall, F. G., D. B. Botkin, D. E. Strebel, K. D. Woods, and S. J. Goetz,  Large scalepatternsof forestsuccession as determinedby remote sensing, Ecology,72(2), 628-640, 1991. Hall, F. G., K. F. Huemmrich, S. J. Goetz, P. J. Sellers, and J. E. Nickeson,Satelliteremote sensingof surfaceenergybalance:Success,failures, and unresolvedissuesin FIFE, J. Geophys.Res., 97, 19,061-19,090, 1992a. Hall, F. G., K. F. Huemmrich, D. E. Strebel, S. J. Goetz, J. E. Nickeson, and K. D. Woods,Biophysical, Morphological, CanopyOptical Property,and ProductivityData From the SuperiorNational Forest, NASA Tech. Memo. TM-104568, 1992b. Holben, B. N., Characteristicsof maximum-valuecompositeimages from temporal AVHRR data, Int. J. RemoteSens.,7, 1417-1434,  1997.  Steyaert,L. T., F. G. Hall, and T. R. Loveland,Land covermapping, fire regeneration,and scalingstudiesin the Canadianboreal forest with 1 km AVHRR and Landsat TM data, J. Geophys.Res., 102, 29,581-29,598, 1997. Tucker, C. J., J. R. G. Townshend,and T. E. Goff, African land-cover classification usingsatellitedata, Science,227, 369-375, 1985.  Tucker,C. J., I. Y. Fung, C. D. Keeling,and R. H. Gammon,Relationshipbetween CO2 variation and a satellite-derivedvegetation index, Nature, 319, 195-199, 1986.  Verhoef,W., Light scatteringby leaf layerswith applicationto canopy reflectancemodeling:The SAIL model,RemoteSens.Environ.,16, 125-141, 1984.  1986.  Jarvis,P. G., J. M. Massheder,S. E. Hale, J. G. Moncrieff, M. Rayment, and S. L. Scott, Seasonalvariation of carbon dioxide, water vapor, and energy exchangesof a boreal black spruceforest, J. Geophys. Res.,102, 28,953-28,966,1997. Justice,C. O., J. R. G. Townshend, B. N. Holben, and C. J. Tucker, Analysisof the phenologyof globalvegetationusingmeteorological satellitedata,Int. J. RemoteSens.,6(8), 1271-1381,1985. Lafleur, P.M., J. H. McCaughey,D. W. Joiner, P. A. Bartlett, and D. E. Jelinski,Seasonaltrendsin energy,water,and carbondioxide fluxesat a northernborealfen,J. Geophys. Res.,102, 29,009-29,020, 1997.  Leblanc,S. G., J. M. Chen,andJ. Cihlar,NDVI directionalityin boreal  F. G. Hall and K. F. Huemmrich, NASA Goddard SpaceFlight Center, Code 923, Greenbelt, MD 20771. (Karl.Huemmrich@gsfc. T. A. Black,Departmentof Soil Science,Universityof British Columbia, Vancouver, B.C., Canada.  P. G. Jarvis,Institute of Ecologyand ResourceManagement,Universityof Edinburgh,Edinburgh,ScotlandEH8 NYL. J. H. McCaughey,Department of Geography,Queen'sUniversity, Kingston,Ontario, Canada.  (ReceivedAugust 16, 1998;revisedJanuary26, 1999; acceptedMarch 4, 1999.)  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items