Hamdan et al (2011) JTFS 23(3)_318-327

of 10
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.
Document Description
Journal of Tropical Forest Science 23(3): 318–327 (2011) Hamdan O et al. REMOTELY SENSED L-BAND SAR DATA FOR TROPICAL FOREST BIOMASS ESTIMATION O Hamdan*, H Khali Aziz & K Abd Rahman Forest Research Institute Malaysia,52109 Kepong, Selangor Darul Ehsan, Malaysia Received September 2010 HAMDAN O, KHALI AZIZ H & ABD RAHMAN K. 2011. Remotely sensed L-band SAR data for tropical forest biomass estimation. Several attempts have been made to obtain forest stand parameters such as stand volume, stand
Document Share
Document Tags
Document Transcript
   Journal of Tropical Forest Science  23(3): 318–327 (2011) Hamdan O et al. 318 © Forest Research Institute Malaysia  REMOTELY SENSED L-BAND SAR DATA FOR TROPICAL FORESTBIOMASS ESTIMATION O Hamdan*, H Khali Aziz & K Abd Rahman  Forest Research Institute Malaysia,52109 Kepong, Selangor Darul Ehsan, Malaysia  Received September 2010  HAMDAN O, KHALI AZIZ H & ABD RAHMAN K. 2011. Remotely sensed L-band SAR data or tropicalorest biomass estimation. Several attempts have been made to obtain orest stand parameters such as stand volume, stand density, basal area, biomass and carbon (C) stocks rom synthetic aperture radar (SAR) data.However the relationship between these parameters and radar backscatter has been a challenging issue sincethe last several years. In this study, L-band ALOS PALSAR satellite image with a spatial resolution o 12.0m was utilised to identiy the relationship between radar backscatter and aboveground biomass o tropicalorest stands. Forest Research Institute Malaysia (FRIM) which has about 420 ha o orest area was selectedas the study area. Field survey was conducted in which 30 plots (50 × 50 m, 0.25 ha each) were establishedand all trees with diameters at breast height (dbh) o 5 cm and above were inventoried. The calculated plot-based biomass was correlated to the pixels o SAR backscatter corresponding to the plot size on the ground.The correlation unction was used to determine stand biomass o the whole study area. Results showed that dense orest was sensitive to the backscatter on horizontal–vertical polarised (HV) image compared withhorizontal–horizontal polarised (HH) image. It was also ound that the L-band SAR backscatter had goodcapability to estimate aboveground biomass in mature stands o tropical orest.Keywords: ALOS PALSAR, SAR backscatter, stand biomass, aboveground biomass HAMDAN O, KHALI AZIZ H & ABD RAHMAN K. 2011. Data penderiaan jauh SAR jalur-L untuk penilaianbiojisim hutan tropika. Beberapa percubaan telah dibuat untuk mendapatkan parameter dirian hutanseperti isi padu dirian, kepadatan dirian, luas pangkal pokok, biojisim dan stok karbon daripada data radarbukaan sintetik (SAR). Bagaimanapun, hubungan antara parameter-parameter tersebut dengan serak balikradar merupakan isu yang mencabar sejak beberapa tahun lalu. Dalam kajian ini, imej satelit ALOS PALSAR  yang berjalur-L dengan resolusi ruang 12.0 m telah diguna untuk mengenal pasti hubungan antara serakbalik radar dengan biojisim atas tanah bagi dirian hutan tropika. Institut Penyelidikan Perhutanan Malaysia(FRIM) yang mempunyai 420 ha kawasan berhutan telah dipilih sebagai kawasan kajian. Bagi tujuan ini, 30plot setiap satunya bersaiz 50 m × 50 m (0.25 ha) telah ditubuhkan dan kesemua pokok yang mempunyai saizdiameter aras dada (dbh) 5 cm dan ke atas telah dibanci. Biojisim bagi setiap plot telah dikira dan dikaitkandengan nilai piksel serak balik data SAR yang bersesuaian dengan saiz plot di lapangan. Persamaan yangterbit daripada hubungan tersebut diguna untuk menentukan biojisim bagi keseluruhan dirian di dalamkawasan kajian. Hasil kajian menunjukkan bahawa hutan yang berkepadatan tinggi lebih sensiti terhadapimej yang berpolar menguuk–vertikal berbanding imej yang berpolar menguuk–menguuk. Didapati jugabahawa serak balik radar berjalur-L mempunyai keupayaan yang baik untuk menganggar biojisim atas tanahbagi dirian hutan tropika matang. *E-mail: hamdanomar@frim.gov.my  INTRODUCTION Forests only cover 28% o the land surace worldwide but contain 80% o the terrestrialcarbon (C), stored as biomass and soil organic C(FAO 2005). Tropical orests are a key component o the global C cycle and contribute more than30% o terrestrial C stocks and net primary production (Wright 2010). However, at the sametime tropical deorestation contributes about one th o total anthropogenic CO 2 emissionsto the atmosphere (Gibbs et al. 2007). In globalenvironment and climate studies, orest biomassis a key variable in annual and long-term changesin the terrestrial C cycle, and is needed inmodelling C uptake and redistribution withinthe ecosystem (Houghton 2005). However theestimations o global terrestrial biomass remainuncertain and are still being studied in line withthe understanding o global C cycle.The suggested schemes or C credit allocationbased on deorestation or C stock baselinesrequire accurate estimates o biomass. Forest biomass can be evaluated using remote sensing   Journal of Tropical Forest Science  23(3): 318–327 (2011) Hamdan O et al. 319 © Forest Research Institute Malaysia  instruments mounted on satellites or airborneplatorms, but substantial reinements areneeded beore routine assessments can bemade at national or regional scales (Bacciniet al.   2004, DeFries et al.   2007). There are noremote sensing instruments that can measureorest biomass directly, thus, additional ground-based data collection is required (Drake et al.   2003, Rosenqvist et al.   2003a). A satellite-basedapproach can provide the spatial pattern o observation needed or biomass at a landscapelevel.For systematic observation at dierent scales,remote sensing is considered a major component o orest monitoring programmes (Lu 2006). Theocus to date, however, has mainly been on theuse o optical data (Toan et al. 2001, Miettinen& Liew 2009). Remote sensing systems that mostly rely on optical data (visible and inraredlight) are urther limited in the tropics by cloudcover (Asner 2001). Meanwhile, research resultsindicate that synthetic aperture radar (SAR) hasa signicant role to play in orest observations(Chenli et al . 2005).The interest in radar remote sensing ormonitoring orest cover rose rom the twoadvantages o SAR data, namely, (1) radar canprovide inormation related to the canopy volume which cannot be produced by other meansand (2) radar has the possibility o acquiringdata over areas with requent cloud cover andnot depending on weather conditions. Theinormation that can be derived rom SAR dataincludes aboveground biomass, annual increment o stand biomass, vertical distribution o biomassand orest stand volume. Rening these estimatesrequires improved knowledge o the densities andspatial distribution o orest C stocks, particularly in high biomass tropical orest ecosystems.Remote sensing methodologies are moresuccessul at measuring aboveground biomassin boreal and temperate orests and in youngstand with lower biomass density (Austin et al.   2003, Rosenqvist et al.   2003b) compared with dense and mature stands. Tropical orestsare among the most biomass- and C-rich but are the most structurally complex ecosystemsin the world that most signals rom remotesensing instruments tend to saturate at certainbiomass level. Short wavelength SAR sensors onboard several satellites such as Earth ResourcesSatellite (ERS-1), Japanese Earth ResourcesSatellite (JERS-1) and Environmental Satellite(Envisat) can be used to quantiy orest biomassin relatively homogeneous or young orests, but the signal tends to saturate at airly low biomasslevels (100–200 t ha -1 ) (Patenaude et al.   2004,Toan et al. 2004, Chenli et al. 2005). Actually,there is potential to improve estimates o biomassacross the tropics or degraded or young orestsbut will be less useul or mature, higher biomassorests (Rosenqvist et al.   2003b, Shimada et al.   2005).Phased Array L-band SAR (PALSAR) onboard the Japanese Advanced Land ObservingSatellite (ALOS), however, promises betterpotential in assessing orest biomass or tropicalecosystem (Baccini et al. 2008). Thus, it hasproduced reliable orest biomass estimates in thisecosystem. Remote sensing oers the possibility o providing relatively accurate orest biomassestimation at a lower cost than inventory studiesin tropical orests; L-band JERS-1 SAR has beensuccessully used in the Malaysian tropical inlandand peat swamp orests (Khali Aziz 2000). SAR isknown to have a response that is directly relatedto the amount o living material with which it interacts. The radar backscatter obtained romSAR is proportional to vegetation density upto a saturation point that is dependent upon wavelength and polarisation o the radar.There are numerous biomass estimationmodels that have been published in the literature.The models vary basically with respect to thecharacterisation o orest and the calculationo its scattering properties. The interaction o SAR signals with the orest is dependent on theradar wavelength and the stand structure o theorest itsel. The requency o SAR is directly proportional to the depth o wave penetration, which means that shorter wavelength can only penetrate the orest or a ew centimetres, whilelonger wavelength can penetrate deeper andsometimes can interact with the orest loor(Imho 1995). Tree elements that play a majorrole in the scattering and attenuation at dierent requency bands are summarised in Table 1.However, the relative contribution o dierent components may change with the tree speciesand their development state. Likewise, at a givenrequency, the scattering and attenuation sourcescan change with the polarisation and incidenceangle.Some SAR systems have the capability to sendand receive energy with dierent polarisations.Since SAR energy can be depolarised uponinteraction with various surace eatures,independently recording the refection o like-   Journal of Tropical Forest Science  23(3): 318–327 (2011) Hamdan O et al. 320 © Forest Research Institute Malaysia  polarised energy (e.g. vertical send–verticalreceive (VV) or horizontal send–horizontalreceive (HH)) and cross-polarised energy (e.g. vertical send–horizontal receive (VH) orhorizontal send–vertical receive (HV)) can yield valuable inormation regarding the characteristicso imaged eatures, and can be particularly useulin the analysis o vegetation type and structure. ALOS PALSAR has two polaristions, HV and HH, which allows polarisation selection.Currently, there is no comprehensive study that uses PALSAR data to map spatial distributiono aboveground orest biomass and C stocks inMalaysia. Thereore, this study was carried out to (1) establish empirical relationship betweenaboveground biomass and PALSAR signalsor tropical orest ecosystem, (2) determineaboveground biomass by using L-band SAR data,and (3) identiy the capability o ALOS PALSAR satellite imagery in estimating abovegroundbiomass. MATERIALS AND METHODSStudy area The Forest Research Institute Malaysia (FRIM)at Kepong, Selangor was selected as the study area. With an area o about 485.2 ha, FRIM issurrounded by the Bukit Lagong Forest Reserve.O the total area, 420.11 ha are covered by orest,o which 379.98 ha are planted orest whichcomprising mostly lowland and hill dipterocarptrees species. The remaining 40.12 ha are naturalorest. Most o the trees planted here are about 80 years old (planted since year 1929). Newly planted trees are also included in the eld datacollection and plot sampling in this study. Figure1 shows the location o study area, marked witha black square. Satellite data  ALOS, an enhanced successor o the JERS-1, was launched rom JAXA’s TanegashimaSpace Center in January 2006. ALOS operatesrom a sun-synchronous orbit at 691 km, witha 46-day recurrence cycle carrying a payloado three remote sensing instruments: (1) thePanchromatic Remote Sensing Instrument orStereo Mapping (PRISM), (2) the Advanced Visible and Near-Inrared Radiometer type 2(AVNIR-2) and (3) the polarimetric Phased Array L-band SAR (PALSAR). ALOS PALSAR image that was used in this study was acquiredon 3rd October 2009. The image came withtwo polarisations, horizontal–horizontal (HH)and horizontal–vertical (HV) and has spatialresolution o 12 m. Field inventory data Field inventory was launched on 22 January 2010and 30 sampling plots o 50 × 50 m size wereestablished within the study area (Figure 2). Theplots cover both types o orests, i.e. natural andplanted orests at various ages. Square plot design was used in this study to acilitate pixel samplingon any satellite imagery, reduce position errorcaused by Global Positioning System (GPS)observation, and make the orientation similarto the shape o the satellite image pixel, whichis normally square. By recording the coordinateo the location only at the centre o the plots(instead o each square edge), position accuracy  was improved. The plot could also be locatedprecisely on the satellite image. By ollowingthis approach, sampling error caused by plot displacement was minimised. While observingthe central coordinate, bearing and distancesrom the central to the quadrate directions weremeasured using prismatic compass and distancetape respectively. Systematic sampling design was applied or the purpose o eld samplingin which selection o sampling plots was basedon orest types and ages. O the 30 samplingplots established, 70% were inside the maturestand orest (o ages more than 50 years) andthe rest were distributed within 5- to 30-year-oldstands. Table 1 Tree elements which are the main scatterers at dierent radar bands or wavelengths Radar bandXCLP Wavelength (cm)2.4–3.753.75–7.515–3030–100Main scatterersLeaves, twigsLeaves, smallbranchesBranches, trunksTrunks   Journal of Tropical Forest Science  23(3): 318–327 (2011) Hamdan O et al. 321 © Forest Research Institute Malaysia  Most o the trees in the study plots are above25 m height and reach up to 45 m, with canopy closure o more than 85% especially or the kapur(  Dryobalanops aromatica  ) trees. These kinds o stands contributed substantially to the level o biomass at specic points and thus introduce variations in the measured biomass in the study area. Associated inormation such as tree height  was also recorded where possible using a laserhypsometer. The dominant species o standingtrees within each plot was also recorded. Most inventories only include trees with diameters at breast height (dbh) ≥ 10 cm but according toChave et al. (2005) and Baccini et al. (2008),about 25% o the total aboveground biomass willbe missed out i the trees with dbh o 5 cm arenot included, which will produce underestimate.Thereore, all trees with the size o dbh ≥ 5 cm were inventoried in this study. Plot level biomass estimation  A tree is made up o several biomass components,namely, oliage, stem, stump, root, bark andbranch. The proportion o these components varies with tree species and tree age. While youngtrees have rather high biomass proportionso leaves/needles and roots, old trees have ahigh proportion o stem biomass. Conversionactors or allometric equation do not provide Figure 1 Location o the study area Figure 2 Layout design o the sampling plot  N50 mNW50 msW50 msENE1 t quarter 35.36 m4 th quarter 2 nd quarter 3 rd quarter Plot center Coordinate: X, Y Tree inside plot Tree outside plot Tree inside plot 
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!