Fan Fiction

16 pages

A lidar-based hierarchical approach for assessing MODIS fPAR

of 16
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
The purpose of this study was to estimate the fraction of photosynthetically active radiation absorbed by the canopy (fPAR) from point measurements to airborne lidar for hierarchical scaling up and assessment of the Moderate Resolution Imaging
  See discussions, stats, and author profiles for this publication at: A lidar-based hierarchical approach forassessing MODIS fPAR  Article   in  Remote Sensing of Environment · June 2008 DOI: 10.1016/j.rse.2008.08.003 CITATIONS 34 READS 78 6 authors , including: Some of the authors of this publication are also working on these related projects: AusCover   View projectRole of forests in global energy and carbon balances.   View projectLaura ChasmerUniversity of Lethbridge 95   PUBLICATIONS   1,461   CITATIONS   SEE PROFILE Chris HopkinsonUniversity of Lethbridge 130   PUBLICATIONS   1,882   CITATIONS   SEE PROFILE Paul TreitzQueen's University 98   PUBLICATIONS   3,003   CITATIONS   SEE PROFILE Harry McCaugheyQueen's University 208   PUBLICATIONS   5,346   CITATIONS   SEE PROFILE All content following this page was uploaded by Laura Chasmer on 18 December 2016. The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the srcinal documentand are linked to publications on ResearchGate, letting you access and read them immediately.  This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institutionand sharing with colleagues.Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third partywebsites are prohibited.In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further informationregarding Elsevier’s archiving and manuscript policies areencouraged to visit:  Author's personal copy A lidar-based hierarchical approach for assessing MODIS  f  PAR  L. Chasmer a, ⁎ , C. Hopkinson b , P. Treitz a , H. McCaughey a , A. Barr c , A. Black d a Department of Geography, Queen's University, Kingston ON, Canada K7L 3N6  b  Applied Geomatics Research Group, Lawrencetown NS, Canada B0S 1M0 c Climate Research Branch, Meteorological Service of Canada, Saskatoon SK, Canada S7N 3H5 d Faculty of Land and Food Systems, University of British Columbia, Vancouver BC, Canada V6T 1Z4 a b s t r a c ta r t i c l e i n f o  Article history: Received 14 April 2008Received in revised form 1 August 2008Accepted 2 August 2008 Keywords: Airborne lidarMODISfPAR LAIScalingBoreal forest The purpose of this study was to estimate the fraction of photosynthetically active radiation absorbed by thecanopy (fPAR) from point measurements to airborne lidar for hierarchical scaling up and assessment of theModerate Resolution Imaging Spectroradiometer (MODIS) fPAR product within a  “ medium-sized ” (7 km×18 km) watershed. Nine sites across Canada, containing one or more (of 11) distinct species typesand age classes at varying stages of regeneration and seasonal phenology were examined using acombination of discrete pulse airborne scanning Light Detection And Ranging (lidar) and coincident analogand digital hemispherical photography (HP). Estimates of fPAR were  󿬁 rst compared using three methods:PAR radiation sensors, HP, and airborne lidar. HP provided reasonable estimates of fPAR when compared withradiation sensors. A simpli 󿬁 ed fractional canopy cover ratio from lidar based on the number of within canopyreturns to the total number of returns was then compared with fPAR estimated from HP at 486geographically registered measurement locations. The return ratio fractional cover method from lidarcompared well with HP-derived fPAR (coef  󿬁 cient of determination=0.72, RMSE=0.11), despite varying thelidar survey con 󿬁 gurations, canopy structural characteristics, seasonal phenologies, and possible slightinaccuracies in location using handheld GPS at some sites. Lidar-derived fractional cover estimates of fPAR were  ∼ 10% larger than those obtained using HP (after removing wood components), indicating that lidarlikely provides a more realistic estimate of fPAR than HP when compared with radiation sensors. Finally, fPAR derived from lidar fractional cover was modelled at 1 m resolution and averaged over 99 1 km areas forcomparisonwith MODIS fPAR. The following study is one of the 󿬁 rst to scale betweenplot measurements andMODIS pixels using airborne lidar.© 2008 Elsevier Inc. All rights reserved. 1. Introduction Covering approximately 76% of the global land surface area,vegetation plays a key role in the functioning of local ecosystemsand can affect processes at scales as large as global weather patterns(Pielke et al., 1998). Leaf area is particularly important, affectingenergyand mass exchangesbetween the terrestrial biosphere and theatmosphere (e.g. Chen et al., 2005). Accurate spatial and temporalestimates of measurable leaf attributes: leaf area index (LAI) and thefraction of photosyntheticallyactive radiation absorbed by the canopy(fPAR), are required as inputs into models of plant production andexchange of heat, water vapour, and CO 2  with the atmosphere (e.g.Chen et al., 2007; Gower et al., 1999). The accuracy of many models,therefore, depends on accurate inputs of these key variables.Chen et al. (2002) de 󿬁 nes LAI as the upper part of the total leaf surface area of all leaves contained within a unit of ground surfacearea(m 2 m − 2 ). Thisde 󿬁 nitionofLAIismostappropriateinthe contextof energy and CO 2 /H 2 O mass balance because it only includes theactively photosynthesizing parts of the canopy and is important forglobal photosynthesis (e.g. gross primary production) modeling usingremote sensing. fPAR can be estimated from radiation sensors basedon the ratio:fPAR   ¼  PAR  AC A − PAR AC ↑   −  PAR BC ↓ − PAR BC ↑    = PAR  AC A  ð 1 Þ where PAR  AC ↓  is the incident PAR above the canopy, PAR  AC ↑  is there 󿬂 ected PAR above the canopy, PAR  BC ↓  is the incident below-canopyPAR after interception by branches and leaves, and PAR  BC ↑  is there 󿬂 ectedPARfromthegroundsurfaceafterabsorptionbysoils(Goweret al., 1999). The MODIS fPAR algorithm, however, excludes PAR absorbedbythe soil (via soil albedo)and onlyconsiders PARabsorbedby vegetation. Despite their importance, LAI and fPAR are dif  󿬁 cult andtime consuming to measure spatially and temporally within ecosys-tems.Bothrequiremeasurementofcanopyfractionalcover(orcanopyclosure) and light transmission (optical methods), or alternativeapproaches involving destructive sampling of leaves and branches.Optical methods are less time consuming than destructive sampling,and are more frequently used ( Jonckheere et al., 2004). These rely on Remote Sensing of Environment 112 (2008) 4344 – 4357 ⁎  Corresponding author. E-mail address: (L. Chasmer).0034-4257/$  –  see front matter © 2008 Elsevier Inc. All rights reserved.doi:10.1016/j.rse.2008.08.003 Contents lists available at ScienceDirect Remote Sensing of Environment  journal homepage:  Author's personal copy temporal measurements from radiation sensors located above andbelow the canopy on a meteorological tower (e.g. Gower et al., 1999;Huemmrich et al., 1999; Schwalm et al., 2006; Eq. (1)). They can alsobe collected spatially below the canopy via incident hemisphericalradiation measurement units such as the LiCOR LI-2000 Plant CanopyAnalyzer, TRAC, and hemispherical photography (HP) (e.g. Chen et al.,2006; Leblanc et al., 2005; Sonnentag et al., 2007). Measurementsfrom radiation sensors on towers are bene 󿬁 cial because they recordphenological changes in vegetation over seasons, but are affected bychanges in solar zenith angles (Hyer & Goetz, 2004). Hemisphericalradiation measurements can also be operated as handheld devices by 󿬁 eld personnel, and can be used at a variety of plots (Leblanc et al.,2005) or transects (Chen et al., 2006) within a larger study area. They are inexpensive to operate, but remote study locations often make itdif  󿬁 cult to measure changes frequently throughout the growingseason (Heinsch et al., 2006). Some comparisons have been madebetween different optical methods. For example, Chen et al. (2006)found that HP tended to slightly underestimate effective LAI ( L e ) byapproximately 8%, on average for a number of forest types whencompared with the LiCOR LI-2000 method.  L e  is related canopy gapfraction estimated using optical methods and assumes that foliage israndomly distributed within the canopy. It therefore does not includethe effects of canopy clumping and may be more associated withprojected leaf areaviewed using remote sensing methods (Chen et al.,2004). Chen et al. (2006) notethat the overall HP estimates of   L e  agreeverywellwiththoseestimatedusingtheLI-2000atanumberofforestsites examined within the Canadian Carbon Program.Other methods used to estimate LAI and fPAR include measure-ments of re 󿬂 ected light collected using remote sensing satellite andairborne platforms (Gamon et al., 2004), for example the ModerateResolution Imaging Spectroradiometer (MODIS). Remote sensingmethods using spectral re 󿬂 ectance alone, however, are not able toresolve the complexity of the vegetation canopy within averagedpixels of fPAR and LAI (e.g. Eriksson et al., 2006; Jin et al., 2007; Xuet al., 2004). Radiative transfer models often improve spectralre 󿬂 ectance measurements by incorporating species-based three-dimensional canopy structure, leaf and stem geometry, and foliagedensity at the tree to canopy level (e.g. Fernandes et al., 2004; Goel &Thompson, 2000; Myneni et al.,1997; Sun& Ranson,2000). These canbe directly related to variability in canopy re 󿬂 ectance measured usingremote sensing methods. However, canopy heterogeneity withinspecies and at different layers within the canopy can lead touncertainties in radiative transfer models (Kotchenova et al., 2004;Tian et al., 2002a,b), and the possibility of numerous results perspecies type (e.g. Koetz et al., 2006). Accurate spatial and temporalmethods of collecting fPAR and LAI would be bene 󿬁 cial and cost-effective for scaling from radiation sensors to wider area coverage.Canopy structural attributes may also be used to better interpretaveraged spectral signatures within lower spectral resolution pixels(Koetz et al., 2006; Kotchenova et al., 2004).The fractional cover of vegetation (used synonymously with fullhemisphere fractional canopy closure in this study), where 1=fullcanopy cover and 0=no canopy cover may also be estimated fromairborne Light Detection and Ranging (lidar). Fractional cover fromlidar may be converted into LAI and fPAR based on the ratio of thenumber of canopy laser returns (single and multiple) to total returns(e.g. Barilotti et al., 2006; Hopkinson & Chasmer, 2007; Magnussen &Boudewyn, 1998; Morsdorf et al., 2006; Riaño et al., 2004; Solberget al., 2006; Todd et al., 2003):  f  cover  ¼  ∑ P  canopy ∑ P  all   :  ð 2 Þ P  canopy isthetotalnumberof laserpulsereturnswithinthecanopy,and  P  all  is the total number of all laser pulse returns within a speci 󿬁 edresolution (e.g. 1 m). Depending on the lidar system used, multiplelaser returns will be recorded from within the canopy and understoryat heights greater than  ∼ 1.5 m above the ground surface (Hopkinsonet al., 2005), but only single returns will be recorded at heights lessthan  ∼ 1.5 m. Solberg et al. (2006) apply a slightly modi 󿬁 ed version of Eq. (2) by including a radiation extinction coef  󿬁 cient. Morsdorf et al.(2006) examine numerous laser pulse ratios and HP annulus ringcon 󿬁 gurations, and found that central rings combined with 󿬁 rst pulsereturns provide the same correlation as when using  󿬁 rst and lastreturns, when compared with  󿬁 eld estimates. However,  󿬁 rst returnstended to yield greater  f  cover than  󿬁 eld estimates, whereas lastreturns tended to yield estimates that were less than those found inthe  󿬁 eld. Also, the extraction of lidar data within a circular areamimicked by HP ( “ data traps ” , Lovell et al., 2003; Morsdorf et al.,2006)werefoundtobemostappropriatewhenradiiofupto2mwereused(Morsdorfet al.,2006). Riañoet al.(2004)relatedlidardatatraps to the height of the tree, and found that this provided the best resultswhen correlating fractional cover to LAI or fPAR. In Hopkinson andChasmer (2007), the use of annulus rings 1 and 2 provided noisyresults because of locally varying canopy gaps and the inability toaccurately geo-register HP to lidar using GPS methods at this scale.They, therefore, opted for a larger data trap of 11.3 m radius andincluded annulus rings 1 – 9. The  f  cover ratio in Eq. (2) may estimate aslightly greater fractional cover when compared with results from HPbasedonannulusringsused,scananglein 󿬂 uences,andtheuseof  󿬁 rstand single returns vs. multiple returns but tends to be within about20% of HP (e.g. Hopkinson & Chasmer, 2007; Morsdorf et al., 2006).Hopkinson and Chasmer (2007) use laser pulse intensity as anindicator of transmission losses through the canopy. They found thatthe intensity-based approach provided slightly better estimates of gap fraction than the commonly used ratio in Eq. (2), but moreimportantly, did not require calibration (e.g. had a 1:1 relationship,regardless of seasonal cycle and sensor con 󿬁 guration).Current studies that use lidar to estimate LAI,  L e , fractional cover(at nadir) or canopy cover (entire hemisphere), gap fraction, and fPAR tend to concentrate on one or a few different forest types within aspeci 󿬁 ed location (e.g. Thomas et al., 2006) and often with controlledlidar surveycon 󿬁 gurations (e.g. Hopkinson & Chasmer, 2007). It is notclearifthereturnratiomethodcanbeapplieduniversallytoarangeof forest vegetation species types and structural characteristics. If themodel is universally applicable (i.e. with little error), then it may beused as a simple methodology for assessing MODIS fPAR and LAIproducts in combination with point estimates from HP. This studypresentsresultsfromacross-CanadatransectoflidarandHPdata.Thereturn ratio method (2) is especially relevant for cases wherenormalized (i.e. geometrically corrected) laser return intensityinformation may not be available (e.g. Hopkinson & Chasmer, 2007,in review). Three objectives will be examined:1. Compare fPAR estimated using radiation sensors and fPAR estimated using HP methods. If fPAR estimated using HP compareswell with fPAR estimated using radiation sensors, then we assumethat fPAR from HP provides a good approximation of fPAR fromradiation sensors, and can be applied spatially.2. fPAR from HP at 486 geo-registered photo plots across Canada arecompared with the lidar fractional canopy cover return ratio(  f  cover) (2).3. Comparisons are then made between 99 1 km resolution MODISpixels of fPAR and pixel-average lidar fPAR within a medium-sizedwatershed.Airborne lidar may provide a useful alternative for mapping fPAR at high resolutions, especially in areas where mixed pixels andunderstory contribute to average re 󿬂 ectance characteristics of lowerspectralresolutionremotesensingproducts(e.g.MODIS)(Serbinetal.,in press). Relationships between mixed pixels, land cover type, andcanopy structural characteristics found in lidar data may be used tobetter understand inconsistencies in MODIS fPAR/LAI products 4345 L. Chasmer et al. / Remote Sensing of Environment 112 (2008) 4344 – 4357   Author's personal copy without the need for extensive  󿬁 eld validation. Inexpensive andsometimes free lidar data are available through a number of websiteand contact listings (e.g. the United States Geological Survey CLICKprogram), providing users with access to already available lidar da-tasets within vegetated environments. This study presents on thehierarchical scaling of point measurements to larger landscape areasusing radiation sensors, plot measurements, high resolution lidar, andlow resolution MODIS pixels of fPAR. If successful, a simple lidarmethodology for estimating fPAR could be an important step towardsimproving ecosystem models and also validating remote sensingproducts, such as those from MODIS. 2. Study areas The study was conducted over 9 sites, alongeast to west and northto south Canadiantransects between the years 2002 and 2007 (Fig.1),with coincident HP. Each site contains one to many different forestspecies types, varying ages, and canopy structural characteristics(Table 1). In many cases, the same species were found at a number of different sites, providing statistical con 󿬁 dence and reproducibility of the experimentindifferent areas. Sitesalsovaryin topography, wheresome sites are  󿬂 at (e.g. Annapolis Valley and York Region), other sitesare gently rolling (e.g. Lac Duparquet and Lake Utikama), whilst stillother sites are mountainous with steep terrain (e.g. Bow Summit andWolf Creek). It is not currently known if topography will in 󿬂 uence thelidar canopy cover; however, inclusion of sites from a wide variety of terrain types will provide some indication of possible errors, if theyexist, as a result of slope angle. Site characteristics and locations areprovided in Table 1 and Fig. 2. Two sites, the Annapolis Valley forest and York Regional forest have been surveyed using airborne lidar andHPmultipletimesthroughoutthegrowingseasonsbetween2000and2007forcontinuingstudiesonphenologyandgrowth (e.g.Hopkinson& Chasmer, 2007; Hopkinson et al., 2008). The White Gull Riverwatershed, which contains the BERMS jack pine chronosequence,Saskatchewan, has been used for MODIS fPAR assessment. Thiswatershed contains a mixture of southern boreal forest vegetationclasses and disturbance regimes, providing an ideal test of bothairborne lidar methods and the MODIS fPAR product. 3. Methodology   3.1. HP data collection and analysis Canopy gap fraction was collected using HP at geo-located siteswithin representative forest types throughouteach studyarea (Fig. 2).Photographic plots were set up in two ways: a) as individual plotscontaining  󿬁 ve photographs. One photograph was taken at the centreof the plot, and four were located 11.3 m from the centre alongcardinal (N, S, E, and W) directions, determined using a compassbearing and measuring tape following Fluxnet-Canada and the Cana-dian Carbon Program protocol (Fluxnet-Canada, 2003); and b) alongtransects of varying lengths and distances between photos. Photo-graphs that were taken within photo plots (a) were located at thecentre of the plot using survey-grade, differentially corrected globalpositioning system (GPS) receivers (Leica SR530, Leica GeosystemsInc. Switzerland; Ashtec Locus, Ashtec Inc., Hicksville, NY) with thesame base station coordinate as was used for the lidar surveys. Geo-location accuracies varied from 1 cm to 1 m depending on the canopycover density at the time of GPS data collection. Measuring tapeand compass bearing methods were then used to locate cardinalphotographs to approximately 1 m to 2 m accuracy. Photographs thatwere taken along transects (b) were located using WAAS-enabled(wide area augmentation service) handheld GPS (Trimble Inc.GeoExplorer, Idaho, USA). These photographs have a locational Fig.1.  Map of the lidar surveys and areas studied. The dark grey area represents the extent and location of the Canadian boreal forest, whereas light grey areas represent southerntemperateforestsand the northern forest-tundratransition. BERMS jackpine sitesare foundwithin the largerWhiteGull River watershed (surveyedat the same timewith the samelidar con 󿬁 guration).4346  L. Chasmer et al. / Remote Sensing of Environment 112 (2008) 4344 – 4357 
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!