Fuel type characterization based on coarse resolution MODIS satellite data

Please download to get full document.

View again

of 5
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
Fuel type characterization based on coarse resolution MODIS satellite data
Document Share
Document Tags
Document Transcript
  Research Article ©    i Forest – Biogeosciences and Forestry    Introduction Fires can cause permanent changes in the composition of vegetation community, de-crease in vegetational covers, loss of biod-iversity, soil degradation, alteration of land-scape patterns and ecosystem functioning thus speeding desertification processes up. Furthermore, fires can contribute to alien  plant invasion, patch homogenization, and create positive feedbacks in future fire sus-ceptibility, fuel loading, fire spreading and intensity. Wildland fires are considered one of the most important disturbance factors in the natural ecosystems of the Mediterranean regions, where every year, around 4!!! forest fires break out causing the destruction of about ".# million hectares $F%& "!!'(. )everal studies $*ila et al. "!!'( dealing with the effects of fires on the vegetation within the Mediterranean basin found that fires induce significant alterations in short as well as long+term vegetation dynamics.revention measures, together with early warning and fast etinction, are the only methods available that can support fire fight-ing and limit damages caused by fires espe-cially in regions with high ecological value, dense populations, etc. n order to limit fire damage, fire agencies need to have effective decision support tools that are able to  provide timely information for /uantifying fire risk. n particular, fire managers need in-formation concerning the distribution, amount and condition of fuels in order to im- prove fire prevention and modelling fire spreading and intensity. 0eographic nform-ation )ystems $0)( and 1emote )ensing $1)( are considered useful tools for support-ing prevention activities $2huvieco et al. "!!4(. 1emote sensing can provide valuable data on type $namely distribution and amount of fuels( and status of vegetation in a consistent way at different spatial and tem- poral scales. )ince the description of fuel  proprieties is usually very comple, fire managers have tried to summarize the phys-ical parameters and spatial distribution of fuel in different classes also known as 3fuel models $%nderson '56", 7urgan 8 1o-thermal '564(. More specifically, a fuel model has been defined as 9an identifiable association of fuel elements of distinctive species, form, size, arrangement, and con-tinuity that will ehibit characteristic fire be-haviour under defined burning conditions9 $Merrill 8 %leander '56:(. ;he spatial dis-tribution of the fuel characteristics can be displayed as fuel type maps. <orthern Forest Fire =aboratory $<FF=( system $%lbini '5:#( is the most common and well+know fuel model that was de-veloped taking the vegetation structure and characteristic of the <orth+%merican floras in to account. 1ecently, in the framework of the rometheus pro>ect $'555(, a new fuel type classification system was specifically developed to better represent the fuel charac-teristic of the Mediterranean ecosystems $http?@@kentauros.rtd.algo.com.gr@promet@ind-e.htm, %lgosystems )%, 0reece(. ;his clas- © SISEF http://www.sisef.it/iforest/ 60 i Forest (2008) 1: 606! Italian National Council of Research (CNR), Institute of Methodologies of Environmental Analysis, C. da S. o!a, I"#$%$% &ito Scalo ('), ItalyCorres*onding Author+ Antonio anorte (alanorteimaa.cnr.it). "itatio# + anorte A, asa*onara R, -%%#. uel ty*e characteri/ation 0ased on coarse resolution M12IS satellite data. iorest 3+ 4%"45 6online+ e0 -#, -%%#7 8R+ htt*+99:::.sisef.it9iforest9 F$el t%pe characteri&atio# 'ase o# coarse resol$tio# *+IS satellite ata ,a#orte A- ,asapo#ara R  A'stract: F$el t%pes is o#e of the ost iporta#t factors that sho$l 'e tae# i#to co#sieratio# for cop$ti# spatial fire ha&ar a# ris a# si$lati# fire rowth a# i#te#sit% across a la#scape. I# the prese#t st$% forest f$el appi# is co#siere fro a reote se#si# perspecti3e. 4he p$rpose is to eli#eate forest t%pes '% e5plori# the $se of coarse resol$tio# satellite reote se#si# *+IS iaer%. I# orer to ascertai# how well *+IS ata ca# pro3ie a# e5ha$sti3e classificatio# of f$el properties a saple area characteri&e '% i5e 3eetatio# co3ers a# cople5 toporaph% was a#al%se. 4he st$% area is locate i# the So$th of Ital%. Fielwor f$el t%pe reco#itio#s perfore 'efore after a# $ri# the ac$isitio# of reote se#si# *+IS ata were $se as ro$#tr$th ataset to assess the o'tai#e res$lts. 4he etho coprise the followi# three steps: (I) adaptation of Prometheus fuel types  for o'tai#i# a sta#ari&atio# s%ste $sef$l for reotel% se#se classificatio# of f$el t%pes a# properties i# the co#siere eiterra#ea# ecos%stes7 (II) model construction  for the spectral characteri&atio# a# appi# of f$el t%pes 'ase o# two iffere#t approach a5i$ lielihoo (,) classificatio# alorith a# spectral i5t$re A#al%sis (4F)7 (III) accuracy assessment  for the perfora#ce e3al$atio# 'ase o# the copariso# of *+IS'ase res$lts with ro$#tr$th. Res$lts fro o$r a#al%ses showe that the $se of reotel% se#se *+IS ata pro3ie a 3al$a'le characteri&atio# a# appi# of f$el t%pes 'ei# that the achie3e classificatio# acc$rac% was hiher tha# 9 for , classifier a# hiher tha# 89 for 4F.;e%wors: reote se#si# *+IS f$el t%pes Tab. 1  + Fuel types classification $rometheus ).*. ro>ect '555(.'0round fuels $cover A!B(grass")urface fuels $shrub cover A#!B, tree cover C !B(grassland, shrubland $smaller than !.D+!.# m and with a high percentage of grassland(, and clearcuts, where slash was not removedDMedium+height shrubs $shrub cover A#!B, tree cover C !B(shrubs between !.# and ".! m4;all shrubs $shrub cover A#!B, tree cover C !B(high shrubs $between ".! and 4.! m( and young trees res-ulting from natural regeneration or forestation;ree stands $A4 m( with a clean ground surface $shrub cover C D!B(the ground fuel was removed either by prescribed burn-ing or by mechanical means. ;his situation may also oc-cur in closed canopies in which the lack of sunlight in-hibits the growth of surface vegetation#;ree stands $A4 m( with medi-um surface fuels $shrub cover AD!B(the base of the canopies is well above the surface fuel layer $A!. m(. ;he fuel consists essentially of small shrubs, grass, litter, and duff :;ree stands $A4 m( with heavy surface fuels $shrub cover AD!B(stands with a very dense surface fuel layer and with a very small vertical gap to the canopy base $C !. m(   Lanorte A & Lasaponara R - iForest 1: 60-64 sification is principally based on the height and density of fuel, which directly influence the intensity and propagation of wildfire $;ab. '(.Eue to the comple nature of fuel charac-teristic a fuel map is considered one of the most difficult thematic layers to build up $eane et al. "!!!( especially for large areas. %erial photos have been the most common remote sensing data source traditionally used $Morris '5:!, Muraro '5:!, &swald et al. '555( for mapping fuel types distribution.  <evertheless, remote sensing multispectral data can be an effective data source available at different temporal and spatial scales that can be fruitfully adopted for building up fuel type maps from global, region down to local scale. For this purpose, up to now, several satellite sensors have been used in last dec-ades. For eample, <&%%+%*G11 $%d-vanced *ery Gigh 1esolution 1adiometer( data were used by Mcinley et al. $'56( for mapping fuel types in western Hnited )tates. =andsat ;hematic Mapper data were used for mapping fuels models in Iosemite na-tional ark, H)% $*an Wagtendonk 8 1oot "!!D(, and in )pain $2ohen '565, 1iaJo 8 2huvieco "!!", )alas 8 2huvieco '554(. % multisensor approach based on )pot and =andsat imager was adopted by 2astro 8 2huvieco $'556( to perform a classification of fuel types for 2hile by using an adapted version of %ndersonKs system. ;he ac-curacies obtained from these researches ranged from #B to 6!B $2huvieco '555(. ;he accuracy level is strongly related with fuel presence and spatial distribution $how many and where( and with specific environ-mental conditions $topography, land cover heterogeneity, etc.(. ;he importance of using multisensor data source to map fuel model was emphasised by many authors $eane et al. "!!'(.%lthough the recognized feasibility of satellite sensors traditionally used for the re-mote characterization of fuel types, the ad-vent of new sensors with improved spatial and spectral resolutions may improve the ac-curacy $2huvieco 8 2ongalton '565( and reduce the cost of forest fire fuel mapping. Hp to now, fire researchers did not paid enough attention to the potentiality of using remote sensing M&E) data to map fuel types and properties.;his research aims to investigate the use-fulness of coarse scale satellite data, such as M&E) imagery, to characterize and map fuel types in fragmented ecosystems. ;his ob>ective is achieved by using rometheus model coupled with M&E) data that were analysed by using Maimum =ikelihood $M=( classifier and )pectral Miture %nalys-is $M;MF( for a test case $located in the south of taly( that is highly representative of Mediterranean like ecosystems. Study area ;he selected study area $Fig. '( is located in the )outh of taly. ;he study area etends over a territory of about 6!!!!! hectares in the 7asilicata and 2alabria 1egions. t con-stitutes a comple morphological unit and it is characterized by comple topography with altitude varies from ! to "D!! m above sea level $a.s.l.( and mied vegetation covers. 7etween ! to #!! m a.s.l., natural vegetation constituted by Mediterranean scrubs and sclerophyllus vegetation is prevailing. From #!! to '!!!+'"!! m a.s.l. the vegetation is  prevailing constituted by etensive popula-tions of Quercus pubescens  and woods of ;urkey oaks $ Quercus cerris (. Eegradation forms are evident, here present as erophytic  prairies and substitution bushes. ;he higher horizons are constituted by beech woods $  Fagus sylvatica ( which arrive up to the '5!! meters a.s.l. of altitude. n the )ila area $mountainous system of region(, between '!!! and '5!! m a.s.l., the more diffuse and important specious is the 2alabrian =aricio  pine $  Pinus Nigra  var. calabrica (. Materials and Methods  ataset !"#$  M&E) $Moderate 1esolution maging )pectroradiometer( is a key instrument aboard the and satellites. ;erraKs orbit around the Larth is timed so that it passes from north to south across the e/uator in the morning, while %/ua passes south to north over the e/uator in the afternoon. ;erra M&E), launched on Eecember '6, '555 and %/ua M&E), launched on May 4, "!!", are viewing the entire LarthKs surface every ' to " days, ac/uiring data in D# spec-tral bands, or groups of wavelengths. ;hese data will improve our understanding of glob-al dynamics and processes occurring on the land, in the oceans, and in the lower atmo-sphere. M&E) is playing a vital role in the development of validated, global, interactive Larth system models able to predict global change accurately enough to assist policy makers in making sound decisions concern-ing the protection of our environment.;he M&E) instrument provides high ra-diometric sensitivity $'" bit( in D# spectral  bands ranging in wavelength from !.4 m to '4.4 m. ;he responses are custom tailored to the individual needs of the user com-munity and provide eceptionally low out+of+band response. ;wo bands are imaged at a nominal resolution of "! m at nadir, with five bands at !! m and the remaining "5  bands at ' km. % N +degree scanning pat-tern at the L&) orbit of :! km achieves a "DD!+km swath and provides global cover-age every one to two days.;he M&E) bands used in this work $;ab."( are the first seven corresponding to a spa-tial resolution of "! and !! m? this choice was performed because these spectral bands are suitable for the study of vegetation char-acteristics. ;he M&E) data used for this study were ac/uired on Ouly, "!!D.%dditionally, photos and air photos were obtained for the investigated area immedi-ately before and after the ac/uisition of satel-lite M&E) data. Fieldwork fuel typing were  performed using a global position system $0)( for collecting geopositional data $latit-ude and longitude(. %ir photos and fieldwork fuel types were used as a ground+truth data- i Forest (2008) 1: 606!61 © SISEF http://www.sisef.it/iforest/ Fig. 1  + )tudy area? $%( =ocation of the study area in M&E) image Ouly "!!D band ' $red(P $7( 107 composition $bands '+4+D( of M&E) spectral channels for the study area. Tab. 2  + Modis spectral bands. BandsLower edge   µm  Upper edge   µm  Piel !e"solution #$% '!.#"!.#:"!"!.64!.66"!D!.4#!.46!!4!.4!.#!!'."D!'."!!!#'.#"6'.#"!!:".'!".'!!   Fuel type c%aracteriation base' on coarse resolution !"#$ satellite 'ata set firstly to identify the fuel types defined in the contet of rometheus system, and secondly, to evaluate performance and res-ults obtained for the considered test area from the M&E) data processing.  Pro(et%eus a'aptation &nly natural vegetated areas were con-sidered for the fuel type characterization. ;he seven fuel type classes standardized in the contet of rometheus system $;ab. D( were detailed identified and carefully veri-fied for the study area on the basis of field works performed before, during and after the ac/uisition of M&E) remote sensing data. n particular, photos and air photos, taken for the investigated region immediately before and after the ac/uisition of M&E) data, along with the fuel types recognized in the field were used for this purpose. Fig. " shows the results obtained from the adapta-tion of rometheus system to the character-istics and properties of fuel types present in the investigated test area. )ignificant patches corresponding to areas representative for each fuel class were carefully identified over the M&E) images by using geo+position data $latitude and longitude( collected during the ground surveys by means of a 0) posi-tioning system. iels relating to these areas were eploited for performing the selection of ade/uate 1egion of nterest point $0round+;ruth dataset( for the seven classes $fuel types( with the addition of ' class con-cerning areas having no Fuel. ;he sample  points of 0round+;ruth dataset were selected in the same areas sub>ect to direct check on field in order to be used firstly to identify the fuel types defined in the contet of rometh-eus system, and secondly, to evaluate per-formance and results obtained for the con-sidered test area from the M&E) data pro-cessing. For this reason, piels correspond-ing to the given 0round+;ruth areas were subdivided in to testing data and training data through randomization of the piels to !B for every class.  !o'el construction an' co(parison ;he mapping of fuel types was obtained by using both a supervised classification based on Maimum =ikelihood $M=( algorithm and )pectral Miture Matched Filtering $M;MF(. ;he M= classifier is considered one of the most important and well+known image classification methods due to its ro- bustness and simplicity. t is wide used in vegetation and land cover mapping. Moreover, it was also tested for fuel model distribution $1iano 8 2huvieco "!!"(. ;he M= method /uantitatively evaluates the vari-ance and covariance of the spectral signa-tures when classifying an unknown piel as-suming at the same time a 0aussian distribu-tion of points forming a cluster of a vegeta-tion class. Hnder this assumption the distri- bution of a class is described by the mean vector and covariance matri which is used to compute the statistical probability of a given piel value being a member of a par-ticular class. ;he probability for each class is calculated and the class with the highest  probability is assigned the piel $=illesand 8 iefer "!!!(. %s above reported, the M= classifier is based on the assumption that dif-ferent variables used in the computation are normally distributed. ;his assumption is generally considered acceptable for common spectral response distribution, but it could be untenable in mied land cover compositions. n this conditions, as piels increase in size, the proportion of mied cover type distrib-uted at piel level will likewise increase and information at the sub+piel level will be of increasing interest. 2onse/uently, in frag-mented landscapes conventional 3hard im-age classification techni/ues provide only a  poor basis for the characterization and map- ping of fuel types giving, in the best case, a compromised accuracy, or, in the worst case, a totally incorrect classification.n these conditions, the use of spectral mi-ture analysis $M;MF( can reduce the uncer-tainty in hard classification techni/ues since it is able to capture, rather than ignore, sub- piel heterogeneity. ;he M;MF allows for © SISEF http://www.sisef.it/iforest/ 62 i Forest (2008) 1: 606! Tab. &  + Fuel type and vegetation typologies adapted from rometheus system for the study area. 'o (uel lowed and bare soilsWoody cultivations)owed lands2alcareous cliffs and detritus)ea, course and water bodies (uel type 1  <atural grassland and pastures (uel type 2 Moors, Hncultivated soils, )ubstitution bushes, )hrubby grassland, 0arigues (uel type & Moors, Hncultivated soil, )ubstitution bushes, )hrubby grassland, 0arigues (uel type ) )clerophyllus vegetationMediterranean shrubs (uel type * 2oniferous forest7eech forest7road+ leaved mied forest (uel type + 7road+leaved mied forest (uel type , 7road+leaved mied forest;ransitional woodland+scrub Fig. 2  + Fuel maps obtained from the processing of M&E) data? $%( Modis 1&P $7( Modis M= classificationP $2( Modis M;MF classification.   Lanorte A & Lasaponara R - iForest 1: 60-64 classifying the proportions of the ground cover types $end+member classes( covered  by each individual piel. Lnd+member classes can be taken from 3pure piels with-in an image or from spectral libraries. &ver the years, different models of spectral mi-tures have been proposed $chku 8 arnieli '55#(. %mong the available models, the most widely used is the Miture ;uned Matched Filtering $M;MF + Garsanyi 8 2hang '554, 7oardman et al. '55, 7oard-man '556( that is based on the assumption that the spectrum measured by a sensor is a linear combination of the spectra of all com- ponents within the piels.n our case, on the basis of ground surveys and air photos, we selected the 1egion &f n-terest $1&( corresponding to the considered seven fuel types, plus ' additional class re-lated to no fuel regions. iels belonging to each of the considered 1& were randomly separated into training data and testing data, used for the both the classification $M= and M;MF( and accuracy evaluation. !esults ;he Fig. " shows the mapping of fuel types obtained for the investigated test area from the M&E) images. )uch map presents very high userKs accuracy. For the accuracy as-sessment we consider the producer accuracy, user accuracy, and overall accuracy, that are defined as follows.;he producerKs accuracy is a measure in-dicating the probability that the classifier has correctly labelled an image piel, for e-ample, into Fuel ;ype ' class given that, on the basis of ground recognition such a piel  belongs to Fuel ;ype ' class. ;he userKs ac-curacy is a measure indicating the probabil-ity that a piel belongs to a given class and the classifier has labelled the piel correctly into the same given class. ;he overall accur-acy is calculated by summing the number of  piels classified correctly and dividing by the total number of piels. Finally, the kappa statistics $( was also considered. t meas-ures the increase in classification accuracy over that of pure chance by accounting for omission and commission error $2ongalton 8 0reen '556(. &verall accuracy is com- puted as the sum of the number of observa-tions correctly classified $class', as class ', class " as class ", etc.( divided by the total number of observations. ;his is e/uivalent to the 3diagonal of a s/uare contingency table matri divided by the total number of obser-vations described in that contingency table $2ongalton 8 0reen '556(.;he ;ab. 4 shows the accuracy coeffi-cients. 1esults from our analyses showed that the use of remotely sensed M&E) data  provided a valuable characterization and mapping of fuel types being that the achieved classification accuracy was higher than :DB for M= classifier and higher than 6DB for M;MF.Hsing the M;MF classification the produ-cer accuracy values improved compared to those obtained from the M= classification for si of the eight considered fuel classes $;ab. (. n particular, the biggest improve-ments were in correspondence of fuel type D $'".:#B(, fuel type # $'D.55B( and no fuel $'4.4DB(. ;he decrease for fuel type " was very low $'.#'B(, while fuel type ' de-creases of '".6" percentage points. ;he user accuracy also increased. n particular, the most significant improvements are for fuel type ' $"6.:B(, fuel type  $'6.D"B(, fuel type : $':.:B( and fuel type $'#.D4B(. ;he decrease for no fuel class is insignificant.%s a whole, results from this preliminary analysis showed that the use of unmiing techni/ue allows an increase in accuracy at around '!B for both the overall accuracy and the appa )tatistic $k( compared to the accuracy level obtained by applying a widely used classification algorithm. )uch results indicate that both the classification and spec-tral unmiing methods can produce reason-ably accurate mapping of fuel type.;he mistakes of classification, their attribu-tion in the different classes and the import-ance of the unmiing, can be verified in the following comparisons? • n M=2 <o fuel class and FtD present a high Miing what consists in a very strong  piel 3transfer $6D piel( from <o Fuel to-ward Fuel type D, due above all to mistakes of selection of the 1&. M;MF, even if not removing the mistake, decreases the 3trans-fer to DD piels and so of beyond #!B. • n M=2, Ft and Ft# present a high 3e-change of piel because the 2lassifier 3moves '5.:DB of the piels attributed in the 0round+;ruth Eataset to Fuel type  to-ward Fuel type # and ""B to Fuel type # toward Fuel type . ;he mistake goes above all attributed what to the difficulty to select distinct 0round+;ruth oints for the two classes, which are very like as regards the vegetational characteristics, being dif-ferent above all on the structural plan. n this case M;MF decreases the mistake of  beyond 4!B. • n M=2 Ft e Ft: show a very high miing that consist in a 3transfer of piel from Ft: to Ft e/uivalent to ":.4"B of the  piels attributed in the 0round+;ruth Eata-set to Fuel type :. %lso in this case the mis-take goes above all attributed what to the difficulty to select distinct 0round+;ruth oints for the two classes, which are very like as regards the vegetational character-istics, for some of the typologies included in two classes. M;MF decreases the 3transfer from ': to # piels with an im- provement, therefore, of almost #B. • n M=2 there is a moderate miing  between Ft: and Ft4 which share some ve-getational typologiesP in the specific case i Forest (2008) 1: 606!69 © SISEF http://www.sisef.it/iforest/ Tab. )  + %ccuracy levels from M=2 and M;MF. -LSSM-LMTMFProducerccuracy #/%Userccuracy #/%Producerccuracy #/%Userccuracy #/% Fuel type '65.:4:'.4D:#.5"'!!.!!Fuel type "6"."##6.!!6!.#:D.DFuel type D#".::D#.#:.D".55Fuel type 45".!#6."55".!#65."DFuel type :#.6:#".!56!.56!.4'Fuel type ##:.D:::.#6'.D#6".:#Fuel type :46.D5#D.6D#.46'.4! <o fuel::."#5:."D5'.#55.## 0erall ccuracy 3 ,&.4& /5appa -oe((icient 3 6.+,+&erall ccuracy 3 4&.+& /5appa -oe((icient 3 6.,72)Tab. *  + 2hange in classification matri. -lassUser cc. ML-User cc. MTMF-hange User cc.Prod. cc. ML-Prod. cc. MTMF-hange Prod. cc. Fuel type ':'.4D'!!.!!"6.:65.:4:#.5"+'".6"Fuel type "#6.!!:D.D.D6"."#6!.#+'.#'Fuel type DD#.#".55'#.D4#".:::.D'".:#Fuel type 46."565."DD.545".!#5".!#!.!!Fuel type #".!56!.4''6.D":#.6:6!.54.!6Fuel type #::.#6".:#."!#:.D:6'.D#'D.55Fuel type :#D.6D6'.4!':.:46.D5#.46.!# <o fuel5:."D5.##+'.:::."#5'.#5'4.4D   Fuel type c%aracteriation base' on coarse resolution !"#$ satellite 'ata the classifier 3moves ''."5B of the piels attributed in the 0round+;ruth Eataset to Fuel type : toward Ft4. M;MF decreases the 9transfer9 with an improvement of bey-ond 4!B. • n M=2 <o fuel and Ft' show a moderate miing with an 3echange e/uivalent to '" piels probably because there is a pres-ence of typical grass species in a few cul-tivated areas. M;MF decreases the mistake to : piels with an improvement of beyond 4!B. • nstead, the improvements in the Miing levels between Ft" and FtD are much less clear. n M=2 between two classes, which, to the Modis spatial resolution can be dif-ferentiated almost eclusively on the struc-tural plan, there is a very high $":( 3e-change of piels. M;MF decreases the mistake only by '6B. ;he same considera-tion can be done for Ft# and Ft: in which the improvement is of "B. • ;he only negative comparison is between FtD and Ft# in which M;MF as regards M=2 provides a worse result of almost 4!B, probably because the Hnmiing un-derlines the mistakes already present in the step of selection of the 1&. -onclusions Multispectral M&E) data were analysed for a test area of southern taly to ascertain how well coarse remote sensing data can characterize fuel type and map fuel proper-ties. Fieldwork fuel type recognitions, per-formed at the same time as remote sensing data ac/uisitions, were used as ground+truth dataset to assess the results obtained for the considered test area. 1esults from our ana-lyses showed that the use of remotely sensed M&E) data provided a valuable character-ization and mapping of fuel types being that the achieved classification accuracy was higher than :DB for M= classifier and higher than 6DB for M;MF.;hus, showing that the use of an unmiing techni/ue allows an increase in accuracy of around '!B compared to the accuracy level obtained by applying a widely used hard classification algorithm. 7oth the classifica-tion and the spectral unmiing methods can  produce reasonably accurate mapping of fuel type. <evertheless, it is more challenging to use the spectral unmiing techni/ues to de-rive fuel type mapping at the subpiel scale.1esults obtained from these investigations can be directly etended to Mediterranean like ecosystems.;he approach proposed in this work can be fruitfully applied to different remote sensed data, such as Quickbird, konos, )&;, %)-;L1, =andsat ;hematic Mapper, or Ln-hanced ;hematic Mapper, <&%%+ %*G11, )&;+*L0L;%;&<, characterized by dif-ferent spatial and spectral resolution for mapping fuel properties at different spatial scale from landscape to regional level. !e(erences %lbini F% $'5:#(. Lstimating wildfire behaviour and effects. 0eneral ;echnical 1eport. H)E% Forest )ervice, ntermountain Forest and 1ange Lperiment )tation <;+D!. &gden, Htah, pp. '+5".%nderson GL $'56"(. %ids to determining fuels models for estimating fire behaviour. H)E% Forest )ervice, ntermountain Forest and 1ange Lperiment )tation 0eneral ;echnical 1eport <;+'"". &gden, Htah, pp. "".7oardman OW $'556(. =everaging the high dimen-sionality of %*1) data for improved sub+piel target unmiing and re>ection of false positives? miture tuned matched filtering. )ummaries of the )eventh O= %irborne 0eoscience Work-shop. O= ublication 5:+', +#. <%)% Oet ropulsion =ab., asadena, 2alifornia, H)%.7oardman OW, ruse F%, 0reen 1& $'55(. Map- ping target signatures via partial unmiing of %*1) data. )ummaries of the Fifth O= %ir- borne 0eoscience Workshop, O= ublication 5+', "D+"#. <%)% Oet ropulsion =ab., as-adena, 2alifornia, H)%.7urgan 1, 1othermal 12 $'564(. 7LG%*L? fire  behaviour prediction and fuel modelling system+FHL= subsystem. H)E% Forest )ervice 0eneral ;echnical 1eport <;+'#:, pp. '"#.2astro 1, 2huvieco L $'556(. Modelling Forest Fire danger From 0eographic nformation )ys-tem. 0eocarto nternational, 'D? '+"D.2huvieco L $'555(. 1emote sensing of large wild-fires in Luropean Mediterranean basin. )pringer+*erlag, 7erlin, pp. '"".2huvieco L, 2ongalton 10 $'565(. %pplication of remote sensing and geographic information sys-tems to forest fire hazard mapping. 1emote )ens-ing of the Lnvironment "5? '4:+'5.2huvieco L, 2ocero E, %guado , alacios %, reado L $"!!4(. mproving burning efficiency estimates through satellite assessment of fuel moisture content. Oournal of 0eophysical 1e-search '!5? E'4+)!:.2ohen W7 $'565(. otential utility of the ;M tasseled cap multispectral data transformation for crown fire hazard assessment. %)1)@%2)M annual convention proceedings? %genda for the 5!Ks. *olume D. 7altimore, Maryland, pp. ''6+'":2ongalton 10, 0reen  $'556(. %ssessing the ac-curacy of remotely sensed data. 212 ress, =ewis ublishers, 7oca 1aton, Florida, H)%.F%& $"!!'(. 0lobal forest fire assessment '55!+"!!!. Forest 1esources %ssessment rogramme, working paper n.  RonlineS H1=? http?@www.-fao.org?6!@forestry@fo@fra@docs@WpTeng.pdf Garsanyi O2, 2hang 2 $'554(. Multispectral image classification and dimensionality reduction? an orthogonal subspace pro>ection approach. LLL ;rans. 0eosci. 1emote )ens. D"? ::5+:6.chku 2, arnieli % $'55#(. % review of miture modelling techni/ues for sub+piel land cover es-timation. 1emote )ensing 1eviews 'D? '#'+'6#.eane 1L, Mincemoyer )%, )chmidt %, =ong E0, 0arner O= $"!!!(. Mapping vegetation and fuel for fire management on the 0ila <ational Forest 2omple, <ew Meico. H)E% Forest )ervice 0eneral ;echnical 1eport 1M1)+0;1+4#+2E.eane 1L, 7urgan 1, van Wagtendonk O $"!!'(. Mapping wildland fuels for fire management across multiple scales? ntegrating remote sens-ing, 0), and biophysical modelling. nternation-al Oournal of Wildland Fire '! $D+4(? D!'+D'5.=illesand ;M, iefer 1W $"!!!(. 1emote sensing and image interpretation. Oohn Wiley 8 )ons,  <ew Iork, H)%.Mcinley 1%, 2hine L, Werth =F $'56(. &per-ational fire fuels mapping with <&%%+%*G11 data. %merican )ociety for hotogrammetry and 1emote )ensing, Falls 2hurch, *irginia, pp. "5+D!4.Merrill EF, %leander ML $'56:(. 0lossary of forest fire management terms. <ational 1esearch 2ouncil of 2anada, 2ommittee for Forest Fire Management, &ttawa.Morris W0 $'5:!(. hoto inventory of fine log-ging slash. hotogrammetric Lngineering D#? '""+'"#.Muraro )O $'5:!(. )lash fuel inventories from :! mm low+level photography. &ttawa, &ntario, 2a-nadian Forest )ervice? #D.&swald 7, Fancher O;, ulhavy E=, 1eeves G2 $'555(. 2lassifying fuels with %erial hoto-graphy in Last ;eas. nternational Ournal of Wildland Fire 5 $"(? D!'+D'5.1iaJo E, 2huvieco L $"!!"(. 0eneration of fuel type maps from =andsat+;M images and auili-ary data in Mediterranean ecosystem. %lcalU de Genares $)pain(, Eepartment of 0eography, %l-calU de Genares Hniversity.)alas O, 2huvieco L $'554(. 0eographic informa-tion system for wildland fire risk mapping. Wild-fire D $"(? :+'D*an Wagtendonk OW, 1oot 11 $"!!D(. ;he H)L of multitemporal =andsat <ormalized Eifference *egetation nde $<E*( data for mapping fuels models in Iosemite national ark, H)%. nterna-tional Oournal of remote )ensing "4? '#D5+'#'.*ila M, =loret F, &gheri L, ;erradas O $"!!'(. ositive fire+grass feedback in Mediterranean.  basin shrubland. Forest Lcology and Manage-ment '4:? D+'4. © SISEF http://www.sisef.it/iforest/ 6! i Forest (2008) 1: 606!
Similar documents
View more...
Search Related
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