download PDF - Imaging Notes

imagingnotes.com

download PDF - Imaging Notes

winter 2004Vol. 19 No. 1THE WORLD’S GUIDE TO COMMERCIAL REMOTE SENSINGParallel processing’sapplications to imagingVirtual carnavigationFire riskassessmentsTimber cruisingefficiencyAnalyzing salmonhabitat


2 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


winter 2004contents » vol. 19 » no.14Cover ImageCalifornia wildfire detail6 MarketScanPeople and Projects831Policy WatchFully implementing the U.S. remotesensing policy2004 Events Calendar6142310121418232629Monitoring Colombia’spipeline infrastructureObserving theft from Colombian pipelinesAudio-visual car navigationFujitsu Ten adds the real world tonext-generation systemsHigh-resolution imageryspawns a new eraMapping riparian vegetation to analyze salmonhabitatA scalable fire riskassessmentMitigating the effects of wildfiresMaking timber cruising moreefficientAutomated timber stand delineation anddensity estimation from aerial imageryMoving toward a paralleluniverseHigh-performance computing for geospatialdata processingNew eye in the sky:RESOURCESAT-1Indian satellite launched.w w w . i m a g i n g n o t e s . c o mW I N T E R 2 0 0 43


cover imageThe image on the cover of the Winter 2004 issue,taken October 28, 2003, is a 1-meter spatial resolution, pan-sharpened IKONOS satelliteimage of Lake Arrowhead, California. Specifically, the area is near the Devore Heights areain San Bernardino County, close to the intersection of I-15 and I-215. Devore Heights is justnorth of the Lytle Creek area and is 12 miles west of Lake Arrowhead.The image was created by blendingthe 1-meter panchromatic band with the4-meter multispectral bands. The image,shot in near-infrared, is displayed as afalse color composite, thereby renderinghealthy vegetation in red-tones andwater in black. Roads, residential areas,docks, and other man-made structures areclearly identifiable in the image, as is landcover type.Of particular interest here are thecloud plumes and hot spots heading upridgefrom the area of San Bernardino onthe north side of I-15, in the lower part ofthe image; north is left in this image. Theimage inset shows several hot-spots nearhomes, and illustrates the detail availablein a high-resolution satellite image. Oncethe fire is contained, the perimeter andseverity of the fire can also be mappedusing satellite imagery. More than 750,000acres were burned in southern Californiaduring this time.The theme of this issue is Natural Resources,and fire risk assessment is coveredin depth. Also, you’ll read how mappingis used to analyze salmon habitat anddelineate timber stands.On other topics, Policy Watch addressescurrent government policy for remotesensing, summarizing progress and opportunitiesfor improvements. Moving Towarda Parallel Universe shows how new technologyin the computing industry will improvethe way geospatial data is processed andexploited, allowing for significantly moreapplication possibilities. «The World’s Guide to Commercial Remote SensingWinter 2004 / Vol. 19 / No. 1PUBLISHERMyrna James YooPublishing Partnerships LLCmyrna@publishingpartnerships.comART DIRECTORJürgen MantzkeEnf ineitz LLCjmantzke@earthlink.net,www.enf ineitz.comEDITORIAL CONTRIBUTIONSImaging Notes welcomes contributionsfor feature articles. We publish articles onthe remote earth imaging industry, includingapplications, technolog y, and business.Please see Contributor’s Guidelines onwww.imagingnotes.com, and email proposalsto editor@publishingpartnerships.com.SUBSCRIPTIONSTo subscribe, please go to www.imagingnotes.com,and click on ‘subscribe.’ Subscriptions are free forthose who qualify.For changes, please submit changes withold and new information toimagingnotes@spaceimaging.com.Imaging Notes (ISSN 0896-7091),Copyright © 2004by Space Imaging LLC,12076 Grant Street,Thornton, CO 80241Imaging Notes is published quarterly byPublishing Partnerships LLC,PO Box 11569, Denver CO 80211Although trademark and copyright symbolsare not used in this publication,they are honored.© 2004 Space Imaging LLCwww.imagingnotes.com4 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


© 2 0 0 4 S P A C E I M A G I N G w w w . i m a g i n g n o t e s . c o m W I N T E R 2 0 0 45


market scanPeopleFormer Governor Joins ESRIFormer Wyoming Governor Jim Geringerhas joined ESRI (Redlands, Calif.).Geringer, a strong advocate for technologyin government, completed his secondterm as governor in January 2003. He alsoserved in the Wyoming legislature for 12years. As a spokesperson for ESRI and theGIS industry, Geringer hopes to spread themessage that GIS technology leads to betterdecision making.www.esri.comRADARSAT International NamesNew PresidentRADARSAT International (Richmond,B.C., Canada) announces the appointmentof Dr. John Hornsby to president. Dr.Hornsby joined RADARSAT International in1991 and has held the positions of directorof Operations, director of WorldwideSales, vice-president of Sales and Marketing,and most recently vice-presidentof Business Development for the upcomingRADARSAT-2 satellite. Dr. Hornsbyhas been involved in both the public andprivate sectors of the remote sensingindustry for over 22 years.www.rsi.caMPT Expands LeadershipTeam with COOMassively Parallel Technologies, Inc.(MPT) (Louisville, Colo.), developer ofhigh-performance computing solutions,hired Dr. Dave Linden as its new chiefoperating officer. He is responsible fordevelopment of MPT’s high performancecomputing production facility in Louisville,where MPT will begin processing seismicdata for oil and gas exploration and aerialimagery at speeds up to 15 times fasterthan conventional supercomputers.www.massivelyparallel.comSpace Imaging NamesNew Sales DirectorsSpace Imaging appointed Jim Roper todirector of Defense/Intelligence SolutionsSales and John Lee to director of Civilianand Commercial Solutions. Roper leadsthe team responsible for expanding SpaceImaging’s business base and sales withinthe Department of Defense and nationalintelligence agencies. John Lee manages ateam of five regional sales managers whoprovide enterprise GIS and imagery solutionsto North American customers.www.spaceimaging.comCompanies and ContractsLeica Geosystems Technology Selected for Australian ProjectsLeica Geosystems (Heerbrugg, Switzerland) announced that its distributor in Australia,C.R. Kennedy, has been appointed sole supplier of surveying equipment for two major transportinfrastructure projects in Australia. The projects include the Western Sydney Orbitalmotorway, the largest and most complex suburban road project in the country, and theChatswood/Parramatta Rail Link tunnel, with 13 km of twin tunnels. The network software isthe Spider GPS reference station software recently launched by Leica Geosystems.www.leica-geosystems.comInta SpaceTurk Grows in theEurasian MarketIn its second year of operation, IntaSpace Systems, Inc.(formerly Space ImagingEurasia) expands its market with newspace sensors and new applications. Thecompany has gained access to SPOT 5satellite in parallel to IKONOS, becoming aone-stop shop for its customers. In Istanbullast year, Jean-Marc Nasr, CEO of SpotImage and S. Murat Erciyes, president ofInta Space Turk signed the SPOT ChannelPartnership Agreement. Inta Space Turkhas also opened a new software developmentoffice at Bilkent University Cyberparkto produce Location Based Services,a fast growing market mainly in the wirelessindustry.Cadastral Solution Implemented byColombian Mapping AgencyESRI (Redlands, Calif.) announces thatits Colombian distributor, Procalculo Prosis,has won a contract with the Geographical InstituteAgustin Codazzi (IGAC) to replace itsexisting automated mapping and cadastresystem. The new system is based on ESRI’sArcGIS software and ArcCadastre, which wasdeveloped by Lantmäteriet, the nationalland survey agency for Sweden, with the helpof ESRI. ArcCadastre is sold worldwide byESRI distributors.www.prosis.comwww.esri.comIKONOS Top 10 Views from AboveEvery January, Space Imaging releases its Top 10 images taken by its IKONOS satelliteduring the previous year. The Top 10 images for 2003 are featured on Space Imaging’sweb site. Tops this year are Baghdad, Iraq during Operation Iraqi Freedom; a 17th-centuryfortress in Bourtange, Netherlands; El Capitan in Yosemite National Park, Calif.; theHong Kong Convention and Exhibition Center; Lake Arrowhead during the devastating 2003California fires; Mount St. Helens, Wash.; the Vatican City, featuring St. Peter’s Basilicaand Square; the Taj Mahal in India; Akkeshi Lake Resort in eastern Hokkaido, Japan; andVictoria Falls, Zimbabwe, one of the ‘Seven Natural Wonders of the World.’IKONOS collects about 1,100 images a day as it orbits 423 miles above the Earth.Since the satellite was launched in 1999, IKONOS has taken more than 1.5 million images,which are stored in Space Imaging’s digital archive.www.spaceimaging.com/gallery/top10_2003/EU Agrees to Work with U.S.Military NeedsA U.S. State Department official reportsthat the European Union has agreedto revise its Galileo global positioningsatellite plan to avoid interfering with theU.S. military’s GPS system. In exchangethe U.S. would provide technical supportto the European effort. Galileo is a 30-satellite constellation navigation tool.European Space Imaging Partnerswith German Space AgencyEuropean Space Imaging (EUSI) in Munich,a Space Imaging Regional OperationsCenter (ROC), has joined a group of fiveother Space Imaging regional affiliatesin the Middle East and around the world.It has partnered with Germany’s spaceagency DLR (Deutsches Zentrum für LuftundRaumfahrt) to run the station.European Space Imaging has the capabilityto collect imagery from above the Arctic Circlein the North to the Sahara Desert in the South,and covers 18 million square kilometers ofEurope, from Portugal to 250 kilometers eastof Moscow. In addition, they have access tothe IKONOS satellite, as a ROC. Customers canpurchase imagery from anywhere in the worldwhile continuing to deal with one company.6 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


ApplicationsIKONOS Satellite LifetimeAnalysis UpdatedIn mid-2003, satellite builder LockheedMartin Space Systems released an updatedlifetime analysis for the IKONOS satellite.The report, developed with nearly four yearsof on-orbit operational data, stated theIKONOS subsystems are expected to lastlonger than originally anticipated. Basedon this evaluation, IKONOS’ life expectancyis estimated to be more than eight years,exceeding the original seven-year designlife. IKONOS is expected to perform at fullperformance until at least the late-2007timeframe if not longer. Historically,spacecraft far exceed their original designlifetimes and can often be operated successfullyfor many years, occasionally atdegraded levels of performance.Tour Mars NowA Mars fly-through on the Web is nowavailable from Skyline Software. The Mars3D model was created from multiple NASAdata sources including imagery from MarsViking Orbiter (global imagery), MarsGlobal Surveyor (high-resolution) and terrainelevation data from the Mars OrbiterLaser Altimeter (MOLA).www.skylinesoft.comIran Earthquake SurveyedThe German Aerospace Center (DLR)is using IKONOS imagery of Bam, Iran toproduce a high-resolution map showingearthquake damage. Widespread damagewas evident throughout the city includingthe historic Citadel. DLR donated themaps to relief organizations.www.spaceimaging.comMonitoring Illicit Crops in AfghanistanThe United Nations Office on Drugs andCrime presented the results of the 2003Afghanistan Annual Opium Poppy Survey inMoscow. UNODC used Space Imaging MiddleEast 4- and 1-meter high-resolution IKONOSimagery to conduct the survey in Afghanistan.The UNODC coupled the satellite imagerycensus with extensive fieldwork that covered28 provinces, 179 districts and 1800 villages.The fieldwork was used to collect ground truthingdata, as well as cultivation information forareas not covered by satellite imagery.The survey estimates the annual turnoverof international trade of Afghan opiates toamount to 30 billion USD. Cultivation covered80,000 hectares in 2003 as opposed to74,000 in 2002.Mohamed El Kadi, managing director ofSIME explained the importance of usingsatellite imagery in illicit crop-monitoring bysaying, “Farmers plant illicit crops in patcheswithin cereals fields. Covered by surroundingcrop, the illicit crop cannot be depicted unlessa very thorough ground survey is conductedwhich may jeopardize the life of the surveyors.Satellite imagery diminishes the hazardsinvolved in conducting such a survey.”www.spaceimagingme.comSpatial Analysis Software Helps India’s Rural PlanningPCI Geomatics (Richmond Hill, Ontario, Canada) spatial analysis software is contributingtowards sustainable development efforts in rural India as a practical facility-planningtool. The pioneering development of a low-cost Geomatics-based FacilityManagement Information System, known simply as “e-Gram Suvidha” (translated as“electronic village facilities”) uses PCI Geomatics GIS technology to help local operatorswho may lack GIS expertise to learn how to perform spatial data queries. The resultingthematic map layers are used for planning decisions at the village level.The e-Gram Suvidha system was developed at the National Informatics Centre (NIC),Ministry of Communications and Information Technology, for the Government of India.The system was designed to utilize PCI Geomatics spatial analysis (SPANS) software, selectedin part for its user-friendly query and analysis capabilities, open-ended design,and high function value.www.pcigeomatics.comReal Estate Development in Dubai Uses ImageryThe Palm, Dubai future landmark and major tourist attraction, usessatellite imagery from Space Imaging Middle East to detect change andmeasure the development of the project at a three-week interval. The Palm,constructed of rock and sand, features a 12 km crescent-shaped protectivebarrier reef extending five km into the Arabian sea. The Palm is composed oflandfill shaped as two palms, each with 17 huge fronds fanning out from thetrunk. In a project that is being executed at a fast pace, satellite imagery isthe best tool, monitoring and mapping the changes in the land surfaces.Collection of images for the first phase began in January 2002, withthe second phase beginning in April 2003. The project was announced inMay 2001, is on fast-track construction, and is expected to host touristsin 2005-2006. At completion, each of the islands will host 50 luxuryhotels, 250,000 villas and apartments, twin marinas, water themeparks, and entertainment facilities.www.spaceimagingme.com «w w w . i m a g i n g n o t e s . c o mW I N T E R 2 0 0 47


policy watchFully implementing the U.S.remote sensing policyFor the overall U.S. satelliteremote sensing industry, 2003 yieldedmixed results. With three companiesstill in business after each experiencedtumultuous commencements, sizeablegovernment contracts for future datapurchases, and a generally supportivenew federal policy, commercial satellitedata suppliers have reason to becautiously content about the near termfuture. With the promise of continuedgovernmental data purchases, thecompanies can be relatively optimisticabout investing in system upgrades forthe long term as well.Data customers, on the other hand,especially many in the scientific community,have suffered two major disappointmentsin the past year. Landsat 7is severely crippled and the governmenthas once again failed to act decisivelyto craft a follow-on strategy that wouldsupport ongoing data continuity for thiscrucial source of high-quality globalenvironmental data.Hence, as we enter 2004, there is stillmuch to do to assure that data customers,the backbone of the imagerycomponent of the increasingly importantgeospatial information industry,are well served and fully supported bygovernment policy. The Administrationneeds to sustain the momentumengendered by its forward-looking 2003policies toward commercial data supplyand also to find the resolve necessary tosupport the policy of Landsat data continuityso clearly spelled out in the 1992Land-Remote Sensing Policy Act.The commercial remote sensing policyreleased on April 25, 2003, representsa significant evolution in the continuingprogress toward sustaining the commercialavailability of high-resolutiondigital data from space. The earlier policy(Presidential Decision Directive 23),crafted in 1994 by the Clinton Administrationprior to any actual experiencewith commercial systems, emphasizedplacing limits on their technical characteristics,such as spatial and spectralresolution, or on data availability duringtimes of conflict. Government use ofthe data was neither encouraged nordiscouraged by the policy, a reflectionperhaps of the intelligence community’sskepticism about the utility of the datafor routine intelligence gathering. As ithas turned out, the commercially-supplieddata have proved highly effectivefor a wide variety of public policy purposes,from disaster response to pro-8 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


viding up-to-date maps of Afghanistanand Iraq. They have also proved usefulin monitoring changes in North Korea’snuclear facilities.The open availability of commercialsatellite data has served the U.S. governmentquite well, as it has been ableto create highly detailed multispectralmaps that can be shared broadly amonggovernment agencies and allies. In thepast such information was generallyclosely held and shared only with a limitednumber of agencies and internationalpartners.Perhaps most important, however,the new policy calls upon governmentagencies to “rely to the maximum practicalextent on U.S. commercial remotesensing space capabilities for fillingimagery and geospatialneeds for military,intelligence,foreign policy,homeland security,and civilusers.” It also directsgovernment agenciesto “focus United StatesGovernment remote sensingspace systems on meetingneeds that cannot be effectively,affordably, and reliably satisfiedby commercial providers becauseof economic factors, civil missionneeds, national security concerns,or foreign policy concerns.”For commercial suppliers and datacustomers, these seem fairly obviouspolicy prescriptions that shouldnot take a formal Presidential policystatement to implement. However, togovernment agencies used to satisfyingtheir data needs by building theirown tailored, one-of-a-kind systems,it seemed a dramatic, revolutionarychange in thought that took yearsand continuing pressure from variousquarters to accomplish. Because of entrenchedagency habits and practices,implementing this policy will continue tobe a challenge in the years to come.The National Geospatial-IntelligenceAgency (NGA—the successor to NIMA, thew w w . i m a g i n g n o t e s . c o mNational Imagery and Mapping Agency)has led the way in implementing this policyby letting major data purchase contractsto industry (the ClearView and NextViewcontracts), which will supply nearly all ofthe routine satellite data needs for theDepartment of Defense and the IntelligenceCommunity. The civil side of thefederal government must now follow suitby aggregating the data needs of thevarious civil agencies under one purchasecontract. Such a move would lead not onlyto reduced data prices for the agencies,but also to more secure funding for thecompanies, enabling them to plan aheadfor system improvements and upgrades.With the new commercial policy, thecurrent White House national securityand space policy team has accomplisheda major step forward. Team members areto be congratulated. Now the governmentshould focus on finding a replacement forLandsat 7 and on creating a continuingsource of Landsat-type data. NASA andthe U.S. Geological Survey have failed toassure data customers, including hundredsof scientists attempting to understandenvironmental change, that highqualitymoderate-resolution data willcontinue to under-gird the nation’s environmentalmonitoring programs. By gatheringdata from most of the world’s landand ice areas in each of the four seasons,Landsat 7 has become the workhorse ofmoderate-resolution systems. Technicalproblems with the scanner mean that itsutility is now severely impaired for many,if not most uses. Unfortunately, there isno replacement system on the horizon,and for most uses, there is no reasonablealternative. Spot 5, the closest operationalalternative to Landsat 7, has morelimited coverage and lacks the spectralrange of the Landsat series.Unfortunately, some government officialsstill have not admitted that Landsatdata have significanteconomic and social valuebeyond that measured in themarketplace. Among otheruses, Landsat data wouldprovide a major pillar ofsupport in the BushAdministration’s internationalinitiative,which was launchedwith great fanfare lastJuly at the Earth ObservationSummit. Involvingmore than 30 countriesand many internationalorganizations, the summitprovided an opportunityfor countries todevelop together a sustained,global programto monitor the world’s environment,and to gather data that can then beused to tackle some of the crucial globalenvironmental and economic problems.In short, despite much progress overthe past year in creating and implementingU.S. remote sensing policy, thegovernment still falls short by not providinga workable plan to continue Landsatoperations. This year’s challenge will be tomake Landsat data continuity a priority.Ray A. Williamson is research professorof space policy and internationalaffairs in the Space Policy Instituteof The George Washington University,Washington, D.C.W I N T E R 2 0 0 49


Monitoring Colombia’spipeline infrastructureObserving theft from Colombian pipelinesThe country of Colombia hasbeen experiencing a problem of theft alongtheir state-controlled (EcoPetrol) gasolinepipelines. Colombia is losing approximately100 million U.S. dollars annually due to thisproblem. They have asked Space Imaging(Thornton, Colo.) to assess the feasibility ofusing the IKONOS satellite to monitor activityalong the gasoline infrastructure.IKONOS imagery, enhanced using techniquesfrom ERDAS Imagine, provides sufficientlyhigh resolution to see within an oilrefinery complex and along roads, railroads,and pipelines. The resolution of the imagery issufficient to identify lineations on the ground,such as rights of way for power lines and scarringof the surface where pipelines have beenburied and run underground. Even with thelimited spatial and temporal data coverageavailable for this project, enhanced IKONOSimagery demonstrated the ability to track thepipeline infrastructure. Therefore, if imageryover the entire petroleum infrastructure wererepeated at sufficiently close time intervals,there is potential to observe where theft is takingplace along pipelines.Two sets of IKONOS images were used inthis study. The first set of imagery consists oftwo scenes of the city of Barrancabermeja, locatednorth of the capital of Bogota. The largestoil refinery in the country is located here,and is capable of processing about 150,000barrels of oil per day (Figure 1).The imagery of Barrancabermeja is a goodstarting point for locating the gasoline pipelinesthat run to and from the refinery, withmost running underground. Space Imaging’sIKONOS imagery provides four-meter colorresolution and, even better, one-meter panchromaticresolution. A multispectral colorimage of the city of Barrancabermeja wasproduced by layering the separate four-metercolor bands of red, green, and blue and thenpan-sharpening the color image to near onemeterresolution.Classifications and enhancements wereperformed on this one-meter color image todetermine the best way to detect the pipelinein and around the refinery center. Many versionsof classifications were performed inERDAS Imagine including supervised andunsupervised techniques. The automatic,unsupervised technique with ten classes in teniterations yielded the best result (Figure 2).For more effective detection and delineationof the pipelines, edge-enhancement algorithmswere applied to the imagery. The goalwas to highlight the edge of the pipelines aboveground and the scarring on the surface due toburied pipelines. Figure 3 shows how one ofthese procedures has helped to define a portionof a pipeline.The strategy used in this feasibility studywas to begin with enhancements and interpretationof the imagery of Barrancabermeja,where above-ground infrastructure is readilyidentifiable, and, using ancillary informationfigure 1figure 2By Tamara Gipprichgeophysicsgraduate student,Colorado School of Mines[http://www.mines.edu],Golden, ColoAknowledgments:Professor Terry Young,Colorado School of Mines,and Mike Nifong ofSpace ImagingFIGURE 1 A close-up of the region of Barrancabermejawhere imagery has been analyzed. Greencoloredregions denote oil fields, while the greenlines are oil pipelines and the red lines are gaspipelines. The two red boxes show the locations ofthe IKONOS images of the city and the refinery.Map courtesy of Wood Mackenzie Global ConsultantsFIGURE 2 A comparison of the pan-sharpenedcolor image and the unsupervisedclassification showing the refinery center ofBarrancabermeja, Colombia. In the classificationimage, the above-ground pipeline,highlighted in yellow, is easily visible.FIGURE 3 A comparison of the color image ofthe Barrancabermeja region and the same imagefiltered by edge detection. Features such asroads, fences and pipelines are sharper in thefiltered image.FIGURE 4 Final interpreted mosaic of Barrancabermejaregion. Red lines denote gaspipelines, green lines are oil pipelines. Dashedlines show where pipeline location has beeninterpolated or extrapolated. Black lines indicatewhere the type of pipeline is uncertain.FIGURE 5 A comparison of the color imageof the remote region and the same imageafter brightness inversion and radiometricbalancing. The pipeline location is evidentin the lower image.figure 210 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


ton, Colo.) IKONOS satellite imagery as themap base. Other systems use simple line mapsto show the car’s location; AVN shows the surroundingbuildings and terrain as they actuallyappear. This means that, in a complex city likeTokyo, the driver can orient himself in a pictureof his actual surroundings.For the first year, over 3,800 square kilometersof image-based maps were provided. These includedthe Tokyo and Osaka metropolitan areas.In 2003 this was expanded to 11,000 square kilometers,with 43 additional cities. In 2004 another6,000 units will be added, with coverage of all theprefectural capitals in Japan.For clear, seamless display of large areas, showingthe imagery as it had been collected was notenough. First, cloud-free, near-vertical imagery ofthe entire area had to be acquired. Next, this imagerywas combined to make a seamless mosaic andwas specially processed for a uniform appearancewith natural-looking color. This makes the preciselocation and appearance information inherent inIKONOS imagery available to the motorist in anattractive color display. A new concept in vehiclenavigation, combining accurate maps and real imagery,had come into being.Since there is a great deal of detail in IKONOSimagery, it is not necessarily optimal for use as aroute guide. Drivers must keep their eyes on theroad and make rapid decisions while moving athigh speed, so a 2D/3D vector map display is alsoprovided. The detailed imagery shows its greatvalue when the driver needs to orient himself ina specific area or to find a particular route, andwhen passengers in the car can enjoy watching areal display of where they are driving. The imageryallows the driver to visualize both present locationand destination in advance from his seat behind thesteering wheel. Graphic icons showing landmarkbuildings, baseball stadiums and many other locationsof interest can be overlaid on the display tohelp the driver find the easiest or best route.According to Mr. Takao Yamaguchi, generalmanager of the Product Planning Departmentat Fujitsu Ten, “IKONOS has emphasizedhigh-resolution Video Graphic Array displaysand enabled more realistic navigation. This hasbrought both convenience and visual enjoymentto drivers and passengers.” The response fromcustomers has been enthusiastic. A 24-year-oldengineer in Osaka said that he frequently uses theIKONOS imagery to orient himself to new locations.Ibaraki states, “It is very convenient whenw w w . i m a g i n g n o t e s . c o mLEFT: A portion of IKONOS sourceimagery, showing seam overlaps, andthe final, specially-processed mosaicof the same location.BELOW: An illustration of the AVNunit, showing a split-screen displaywith image and vector maps of thesame location.I drive in an area where I am not familiar with so much.” A 29-year-old lady onan office staff in Hyogo Prefecture explained, “It is good to see the real appearanceof the surroundings.” There is also the attraction of new technology. Anengineer in Osaka said, “It’s fun! I can impress friends with IKONOS.”It has been a long and complex process to bring high-resolution satellite imageryto CNS. The very detailed imagery requires a great deal of storage, evenwhen the data are compressed. Systems which use only vector maps can storethe data on CDs. The first generation IKONOS AVN8802 used DVDs; in thesecond generation AVN9902 the DVDs have been replaced by two removable20GB hard disk drives (HDD). Fujitsu Ten will add new coverage to the HDDsas it becomes available.Japanese CNS manufacturers are reaching a common conclusion: CNS shouldprovide accurate, easy-to-use route planning and driving information, and shouldalso serve as multi-function entertainment centers. With AVN, the CNS has progressedfrom a relatively simple driving tool to a partner to the driver, providingnavigation, location information and entertainment.The CNS market continues to grow rapidly, and drivers are demanding moreand more capabilities and features from their navigation systems. Fujitsu Ten isworking to stay ahead of customer expectations by adding more coverage area,display zoom, additional content and other enhancements. In all of this growth,satellite imagery is expected to play an important role. Image-based displays willremain a major discriminator among CNS options, and Fujitsu Ten aims to keeptheir competitive edge by providing various kinds of critical information to driverswith maximum practical value. «W I N T E R 2 0 0 413


1Riparian vegetation provides a number of importantfunctions to the ecosystem. These include adding nutrients to streams fromlitter fall, bank stabilization by the root masses of trees within the area,shade which controls water temperature and large woody debris for streamchannel development.One of the goals of the Oregon Forest Practices Act is to maintain riparianareas in mature forest conditions that provide large conifers over time. Thus,riparian zones may provide much of the older forest on private forestland.While society’s strategy for providing fish and wildlife values rests heavily onriparian vegetation, we do not have a good inventory of the existing conditions.Many ground-based samples have been designed to avoid riparian areas or havenot specifically noted which inventory plots were taken in riparian areas. Aslocal watershed councils and landowners plan to improve aquatic habitat conditionsfor various species such as coho salmon, they need to know about the conditionof the riparian vegetation in the watershed. For example, decisions abouthow to improve water temperature, where to place woody debris to increasepool habitats, or where the most effective refuges might be located are decisionsthat would benefit from knowledge about riparian vegetation.Most of the consistent information describing large areas of forest vegetationhas been produced from satellite imagery because it is the most cost-effective wayto inventory large areas. One of the key satellites used in mapping vegetationthroughout the state of Oregon has been Landsat. The Landsat satellites collectspectral data for individual pixels that are approximately 30 meters in size,roughly equivalent to a baseball diamond. While Landsat is an excellent, cost-ef-High-resolutionimagery spawns23a new eraByMapping riparian vegetation to analyze salmon habitatKevin BirchResource Policy Analyst,Oregon Department ofForestry, Salem Ore.4Rick JonesDirector NorthwestOperations,Space Imaging,Portland Ore.Jim MuckenhouptRemote Sensing Analyst,Space Imaging,Portland Ore.14 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


fective platform for regional mapping, the sizeof the pixel presents difficulty when attemptingto map riparian vegetation. Riparian zones aretypically diverse areas, and in many areas areno wider than 30 meters. Thus, detailed classificationsof riparian vegetation types fromLandsat are difficult.With the growing constellation of highresolutionsatellites, leveraging automatedprocesses to map riparian vegetation types hasbecome feasible. The Oregon Department ofForestry undertook a project to evaluate theeffectiveness of using higher resolution satellitedata to map riparian vegetation. The project,which is currently in process, was developed toevaluate three different image datasets for boththeir spatial/spectral capabilities to map riparianvegetation and their cost-effectiveness to doso. The datasets utilized include:a. Landsat 25m Multispectral mergedwith IRS-C 5m Panchromaticb. IKONOS 4m Multispectralc. IKONOS 1m colorSTUDY AREAThe study area for this project was thefifth-field watershed in the Yaquina RiverBasin predominantly covered by the Elk Cityand Harlan 7 1/2 minute U.S. GeologicalSurvey quadrangles (Figure 1). It is approximately57,000 acres in size and contains 235miles of streams on forest and agriculturallands. The watershed is predominantly publiclyowned with about 73 percent of the landowned by the federal government, 13 percentowned by the state and only 14 percent ownedby private landowners. The majority of thewatershed is zoned as forestland and only sixpercent of the land is zoned as agricultural.However, for riparian protection purposes,this small amount of agricultural land isvery important because it is made up of longnarrow pieces that follow the river bottoms.Although the agricultural land makes up onlysix percent of the land, it contains 27 percentof the streams that have been identified by theOregon Department of Fish and Wildlife asimportant for coho salmon.CL A SSIFIC ATION SYSTEMFor this project we used a classificationsystem that consists of three distinct attributes:cover type, size and canopy closurew w w . i m a g i n g n o t e s . c o m(Figure 5). The purpose for generating three layers instead of a single layer was to facilitate differingcombinations of these variables so that the data could be easily cross-walked and compared todifferent classification schemes developed by both the state and the federal governments.PROCEDUREMapping riparian vegetation, rather than sampling,has the advantage of allowing us to ask specificCover Conifertype Hardwoodquestions about the geospatial location of features ofConifer/Hardwood Mixinterest. The amounts and locations of different riparianconditions are critical to answering basic questionsBrush/Recent ClearcutCultural Stringersabout effects of different land-use practices on the landscapeand how the landscape may change over time. For 5Non-ForestWaterexample, a recent clear-cut will over time grow backCanopyinto forest. However, land managed for agriculture will< 10 percentlikely remain agriculture without a change in management.Understanding the differences enables us to mod-40-75 percentclosure 10-40 percentel the recruitment of large woody debris into the future> 75 percentby running different types of simulation models.Size < 1” DBHTo create the map, Space Imaging classified each> 1 - 8” DBHimage source independently, differentiating between> 8 - 21” DBHupland and riparian areas by means of a buffer. The> 21 - 39” DBHbuffer width was chosen to be 500 meters. This width> 39” DBHcorresponds to areas where there is generally riparianvegetation. While the Oregon Forest Practices Act wasdesigned to provide increasing protection as the streamsize increases, up to 100 feet from a stream, this largerbuffer size was selected to include more area to maintain flexibility in further policy analyses.Once the buffers were generated, Space Imaging derived polygons from the imagerybased on a minimum mapping unit of 0.5 hectare because of the spectral and texturalsimilarity of the features in the imagery. Next, a hybrid supervised/unsupervised classificationmethod was utilized to classify the cover type, size and canopy closure layers. As withmost natural features, a certain amount of heterogeneity is inherent on the landscape andthus needed to be accounted for in the classification. While our minds tend to clump similarfeatures (i.e, develop polygons around features), image processing tends to split features.To account for these phenomena and to label areas where features are considered similarenough to be in the same class, decision rules were generated to group the cover type, sizeand canopy closure variables into the discrete variables of the classification scheme. Thisinformation was then used to generate the maps illustrated in Figures 2 through 4.DISCUSSIONMaintaining riparian vegetation is an important part of the strategy to produce and maintainfish and wildlife habitat. Thus the development of policy options to manage these areasrequires a thorough understanding of the condition of these areas over time. Unfortunately,the large landscape-scale data sets that are currently available do not seem to adequately describeriparian vegetation; thus the use of higher resolution data sources is required.While this project is in process, preliminary results illustrate that higher resolutionsensors are capable of identifying the vegetation types within diverse linear features suchas riparian zones. Once the classification of the vegetation types is complete, an accuracyassessment will be conducted in conjunction with the Pacific Northwest Research Stationof the USDA Forest Service. This assessment will provide valuable insight into how well theinformation derived from the imagery coincides with the vegetation on the ground. This informationwill provide managers and policy makers with key information which can be usedas decision criteria on a platform that can be used under various policy considerations. «W I N T E R 2 0 0 4 15


2fire risk andhazard assessmentframeworkBy Donald CarltonPresident,Fire Program Solutions LLC,Estacada, Ore,www.fireps.comJames L. SmithDomain Sales Manager,Forestry, Fire and EcosystemsManagement,Space Imaging LLC,Thornton, ColoJulie CoenSenior Project Manager,Space Imaging,Salt Lake City, UTThe news is replete with images of wildfire, among them theGrand Prix and Cedar Fires in Southern California, the Biscuit Fire in Oregon, and the HaymanFire in Colorado (Figure 1). The loss of millions of acres and thousands of homes hassignificantly affected countless lives. Because of the devastating economic and human impactsof these events, there is high value in addressing how to mitigate the effects of wildland fires.One of the most important and proactive ways to identify areas most likely to be impactedby wildland fires is to perform a wildland fire risk assessment. After a wildland fire risk assessmentis complete, planners and fire professionals can identify the locations of likely impacts,and can analyze the value of mitigation measures. With accompanying analytical tools,the results of a wildland fire risk assessment can be used in simulations to determine whatactivities have the highest potential benefit, such as adding another suppression resource in aspecific location, or implementing a fuels treatment program. Wildland fire risk assessmentscan also serve as a baseline for monitoring change in fire susceptibility and effects over time.Assessments can be conducted at different scales. The appropriate assessment scaledepends on the decisions individuals are trying to make. At one end of the spectrum arelandscape scale wildland fire risk assessments. These can be followed by community levelprioritization and assessments resulting in mitigation options based on either scale. Assessmentis frequently used to prioritize fire effects in Wildland Urban Interface (WUI)areas. Each assessment type is a component of an overall wildland fire risk mitigationstrategy, and it is important that these components be consistent and compatible acrosstime and space. Temporal and geographic compatibility in the wildland fire risk assessmentapproach is important to allow for comparability between areas in support of cost-efficiency.In this article, a wildland fire risk assessment approach is presented which is compatible over timeand geography. The details upon implementation may change, but the overall framework for the assessmentis consistent. This approach has previously been applied in the State of Florida, and is beingused currently in the Southern Wildland Fire Risk Assessment (SWFRA) project.GENER AL RISK /HA Z ARD FR A MEWORKA general approach to wildland fire risk assessment was developed by the Florida Division of Forestryand Space Imaging. The project deliverables included a wildland Fire Risk Assessment (FRA) documentingthe current situation. An ArcView extension called the wildland Fire Risk Assessment System (FRAS) wasthen developed to allow users to view the current situation data layers and also to model the effectivenessof some mitigation options. The wildland fire risk assessment framework was originally developed byDonald Carlton of Fire Program Solutions LLC, who was employed by Space Imaging as a consultant forthe project. The methodology used was based on the Lake Tahoe Fire Risk Assessment completed for theUSDA Forest Service in 1999 by Donald Carlton and Dr. Mark Finney.18 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


Mitigating the effects of wildfires1w w w . i m a g i n g n o t e s . c o mW I N T E R 2 0 0 419


345In this article, several words will be used in a specific context. Frequently, fire professionalswill define what starts fires as “risk.” Fire ignition potential will be referred to as fire occurrence,not risk. In the same context, fire professionals frequently refer to what burns as “hazard.” Also,the fuels that will burn in a wildland fire will be referred to as fuels, not hazard. Lastly, fire professionalsfrequently refer to what the effects of a fire will be as “values at risk.” Values that couldbe affected by a wildland fire will be referred to as values effects, not values at risk. Webster’sdictionary defines risk as “the possibility of suffering harm or loss.” The analytical integration ofthe likelihood of a wildland fire, the potential fire behavior, the success of fire suppression forcesgiven the fire behavior and the resultant fire’s effects will be defined as “risk” in this article. Thisuse of the term risk is consistent with its accepted definition.The wildland fire risk assessment framework is presented in four columns (Figure 2). The firstcolumn of the model represents Compiled Inputs, and largely consists of locating or creating datasets that form the basis for the wildland fire risk assessment. This is the most critical, and in someways, the most difficult step in the wildland fire risk assessment process. Fuels can be particularlydifficult and expensive to quantify because they are typically geographically diverse and highlyspecific. It is common for some type of imagery to be used in the development of fuels data, withthe specific imagery types and processes depending on the scale of the assessment. Another criticaldata set is a listing of historic wildland fire ignitions based on past fire occurrence records. Fireoccurrence data can be difficult to gather and aggregate, primarily when non-federal governmentfire protection agencies are involved.Two well-defined wildland fire reporting systems are used by the five federal wildland fireprotection agencies (USDA Forest Service, the USDA Bureau of Land Management, the Bureau ofIndian Affairs, the National Park Service and the Fish and Wildlife Service). Fire reporting systemsused by each of the 50 states are highly variable. In addition, large tracts of land can be managedby other agencies such as the Department of Defense where there is no common fire reporting format.Other data needed includes infrastructure (roads, resource locations), terrain, forest canopyclosure, historic weather, wildland fire suppression capabilities, historic fire suppression costs, andvalues affected. The compiling of the necessary data sets needed to quantify wildland fire risk isnot a trivial process, and presents unique challenges.The second component of the wildland fire risk assessment framework is Derived Outputs.In this step, information is derived from the basic data using ranking systems, analyticaltools and predictive models. Spatial analysis methods are used to create “Fire OccurrenceRates” from raw historical data on fire ignitions in the local area.For example, specific weather conditions are generated within derived zones of statisticallysimilar weather and input into a fire behavior model called FlamMap. FlamMap is a set ofraster prediction models developed by Dr. Mark Finney. FlamMap uses information on slope,aspect, elevation, wildland surface fuels, and canopy closure, as well as aerial fuels components,together with specific weather conditions to calculate fire behavior values for each cell on thelandscape. Specific relationships are developed between fire spread rate and the expected finalfire size given a defined level of fire suppression effort. Indices are developed to measure the effectof fire suppression costs and physical fire effects.The development of Modeled Indices is the third major component of the wildland fire riskassessment framework. In this portion of the process, indices for Fire Response, Wildland FireSusceptibility and Fire Effects are created from the various Derived Outputs. The goal of thisstep is to generate ranking information that can be combined into a simple set of outputs fordecision makers. The Wildland Fire Susceptibility Index (WFSI) is related to the probabilitythat a specific location will burn in a wildland fire event. WFSI combines information from thefire occurrence rates, the modeled fire behavior based on weather conditions and rate of spreadversus final fire size relationships (Figure 3). The Fire Effects Index captures information fromthe fire suppression costs and the physical fire effects.The final step in the wildland fire risk assessment process is the generation of products tosupport fire-planning efforts. Figure 2 lists some of the typical products from a wildland firerisk assessment, including compiled data sets, and standard maps and reports for fire planners20 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


and the public. Intermediate information can also be important, such as estimates of ignitionpotential, maps of fuels, maps of structural hazards, the identification of critical values orresources that might be effected by a fire, and an assessment of the effectiveness of the firesuppression organization. In many cases, the final outcome of a wildland fire risk assessmentis some type of mitigation plan that aims to proactively reduce the number of, or effects fromwildland fires. A well-designed wildland fire risk assessment provides all the informationneeded to prioritize areas for analysis of potential mitigation measures.SC AL ABILIT Y OF THE WILDL AND FIRE RISK A SSESSMENT FR A MEWORKThe scale of a wildland fire risk assessment should be driven by goals and objectives. Is thegoal to develop a strategic plan, or a detailed tactical plan? Regardless, the basic wildland firerisk assessment process is the same across geographic and temporal scales. Data is compiled,values are computed or derived, and information is ranked and combined to create final outputs.However, the scale (objective) of a wildland fire risk assessment typically impacts theexact type and resolution of the data required.The impact of scale, and the scalability of the wildland fire risk assessment frameworkdescribed in this article can best be demonstrated through specific examples. The uniqueaspects of three different scales of risk assessment will be presented, beginning with thecoarsest scale analysis and progressing to the most detailed.6community with a 3 mile bufferLandscape Scale AnalysisThe Florida Division of Forestry enlisted Space Imaging to conduct a state-wide wildlandfire risk assessment on an area of more than 35 million acres. The goals of this projectwere to create a database and set of application tools that increase the understanding ofwildland fire risk across the state, identify the areas of highest concern across the state, andassist efforts by the State of Florida to create a strategic fire mitigation plan.Fuels are often the most critical and costly data set, and thus tend to dictate the designand resolution of the remaining risk assessment data sets. In the Florida project, fuels weremapped using Landsat 7 (Figure 4), which dictated a 30-meter resolution for the remainingraster data sets and the final project results (Figure 5). Higher resolution imagery wouldhave been more costly to obtain, difficult to store for an entire state, and unwieldy to use.The coarser 30-meter resolution made sense for a strategic, regional analysis. Informationon specific structure locations or types, defensible space, or other structural hazards is notneeded for a project with this objective, eliminating that component of the framework and itsassociated analyses. However, information on basic infrastructure and the location of housingdevelopments was needed to assess the potential fire effects, although at a lower level of detailthan would be needed to develop a mitigation plan for an individual community.7community zone WFSIRating CommunitiesIn the real world, there will very likely be many more communities seeking financial assistancefor detailed planning than can be supported by the available resources, creating theneed to rate or rank the candidate communities. The process of prioritizing communitiesbegins with a landscape scale analysis, and adds some new data elements and some spatialanalysis components, each of which is contained within the fire risk framework illustratedin Figure 2.The most important new data element is the locations of communities to be ranked.More detail or more accuracy may be needed than in a landscape analysis because thegeographic focus has narrowed to individual communities from the larger surroundinglandscape. Communities can be located as a “point” on the landscape, or a polygon if theboundary information is available. Point features may be sufficient for small communities,in which case a circular buffer is generated to create an area for the feature. Polygonalboundaries are needed for larger communities with irregular shapes, and a buffer can beadded using standard GIS software tools (Figure 6-8).w w w . i m a g i n g n o t e s . c o m8community A zone average WFSI ratingW I N T E R 2 0 0 421


The next analytical step is to overlay these community boundaries with the landscapescale wildland fire risk assessment to determine the situation in and around the communities,in what is commonly called the Wildland Urban Interface. Various mathematical rasteranalysis procedures can then be used to summarize the situation within the community andthe WUI. The details of the raster analysis will likely vary for different clients or in differentregions of the county, but this process is objective and repeatable, and could be rerunas the community boundaries change through time and give consistent, comparable results.The final result is simply a rated set of communities. These community ratings are then usedwith other information in a prioritization or ranking process by local decision makers todetermine how resources will be allocated.roof type mappinghazard proximity analysis910Community Hazard AssessmentsThe most detailed wildland fire risk assessments take place at the individual communitylevel, and are often called community level assessments. This intensive level of assessmentmay only take place on the communities identified as highest priority after the rankinganalysis because of the cost of data collection. At this level, the focus of the analysis is onhuman constructs (structural hazards), and often requires a combination of high-resolutionimagery and field-collected information.Important structural hazard assessment information often can be generated quickly andcost-effectively on imagery. Automated image processing techniques combined with highresolutionsatellite imagery such as the natural color IKONOS image can be used to identifyand map rooftop footprints more quickly than field visits (Figure 9). Roof types for thesestructures, such as asphalt shingle, clay tile or wood shake, have also been successfully identifiedon high-resolution satellite imagery, which may be a critical element of a structuralhazard assessment. Distance from the structure to flammable vegetation, also known as“defensible space” can also be derived on high-resolution imagery by combining vegetationlayer and structure layers in a GIS (Figure 10).The remotely sensed information may need to be combined with data collected in the field tocomplete the community level assessment. Exterior flammable structures such as decks or outbuildingsunder vegetation, flammable exterior walls, or unique situations at individual structures,may need to be identified in a house-by-house field visitation. The key is to utilize remote sensingand field data collection optimally in combination to reduce the total cost of data collection.Once all the required data are compiled, the process outlined in the wildland fire risk assessmentframework is again followed — derived information, modeled indices and outputs. Thefactors affecting the overall ignition potential-fuels, fire behavior, fire suppression resources effectiveness,and fire effects-are the same as in a landscape level wildland assessment. The levelsof detail or specific analyses may be different. The output of a community level assessment canvary widely, however. Whereas the focus of a wildland fire risk assessment is usually on thewildland fuels and fire ignition potential, a community hazard assessment is focused on the interactionbetween wildland fuels and structural hazards. Community level assessment requiresvery specific information based on the decisions to be made.SUMM ARYWhile they may appear to be unique on the surface, different types of wildland fire risk assessmentshave a common kernel that can be used to define the processes. A general wildland fire riskassessment framework can be used to: 1) Conduct a landscape scale wildland fire risk assessment, 2)Rank communities according to the likelihood an acre will burn, and 3) Support a community levelassessment and the development of a community-specific mitigation plan. Because wildland firerisk assessment is a spatio-temporal phenomenon, it is important to have a process that is flexibleand consistent over time and space. It is also critical that the process use proven analytical processesbased on accepted science to integrate adequately the inputs to the intermediate and fine measuresof potentials for effects. Planners and fire professionals need decision support tools that give reliableand valid answers, especially for questions relating to the efficiency of fire mitigation programs. «22 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


y Matt VernierSenior Remote Sensing Analyst,Space Imaging Solutions,Ann Arbor, Mich.Portland, OregonAutomated timber stand delineation anddensity estimation from aerial imagery‘Cruising Timber’ is one of the primary jobsof a field forester, either in private industry or public naturalresource agencies. The purposes of timber cruising are toestimate stocking levels and timber volumes, and to gaugeoverall forest health with regard to natural resource management.It is a necessary part of the commercial timberindustry, and is also very time-consuming and expensive,especially when land holdings are large and dispersed.Such is the concern of Plum Creek Timber Co. Inc. (Watkinsville,Ga.), the second largest private timberland ownerin the United States, with eight million acres located in thenorthwestern, southern, and northeastern regions of thecountry. To address this concern, Space Imaging (Thornton,Colo.) is developing a method for mapping timber stands usingaerial imagery. Plum Creek had already collected digitalaerial imagery (Emerge, Andover, Mass.), over much of theirlandholdings in Mississippi at a 2-meter pixel resolution ina false color format (Near-Infrared, Red, Green) (Figure 1).The challenge is to come up with a method that is both accurateand cost-efficient.Plum Creek cruises all plantations at age 12. To reduceinventory costs while maintainingsurvey precision, Plum Creek plansmoreto do a type of stratified cruisebased upon criteria such as sitequality and stocking class. Estimatedsite quality information is available from other sources,but there is no reliable estimate of stocking of 12-year-oldstands. If personnel travel to each stand to obtain an initialstocking estimate, they might as well go ahead and conductthe cruise since a significant portion of the total cost of astandard cruise is travel to the stand.Makingtimbercruisingefficientw w w . i m a g i n g n o t e s . c o mW I N T E R 2 0 0 423


1Portion of Emerge aerial imagery obtained over Plum Creek’sownershipIn addition, stand boundaries are often inaccurate due tovarious factors, such as:a. Intrusion of other species near the edges of a standsince the last map updateb. Excessive hardwood competition or pine seedlingfailurec. Inability to identify the actual stand boundary duringthe early years of a plantationd. Errors in mapping during the initial stand identification.Accurately identifying a stand boundary is important notonly from a location standpoint, but also because it impacts theaccuracy of the stand area estimate. Cruise information is scaledby the estimated area of the stand, making the quality of the areaestimate as important as the quality of the cruise information.LOOKING FOR A SOLUTIONSeveral software packages for stand delineation and stockingidentification were evaluated. The software package thatshowed the most promise for automating the process was eCognition(Definiens GmbH, Munich, Germany) with its capabilityto treat imagery both as a raster and vector dataset. eCognitionis designed to segment the image into units of similar spectraland spatial patterns and to classify those segments according toa pre-defined rule base. Importing and segmenting the imageryat multiple resolutions, and exporting the resulting standdelineation as a shapefile can be automated, significantly reducingthe amount of time and manpower devoted to the task,when compared to manual delineation.The approach taken to identifying stands of pine consisted oftwo stages. In the first stage, the individual pines or pine clustersare classified using very small, or base-level segments (Figure 2).In the second stage, using higher-level segments, each with a muchlarger area, the relative area of pines is calculated. If the relativearea of pine is high enough, the segment is classified as a pinestand. Those that are classified as pine stands can be subdivided,again using the relative area of pine, into different stocking levels(Figure 3, legend shown in Figure 4). Implementing this approachin eCognition makes use of the multi-resolution capabilitiesof the software. In higher levels of segmentation, the polygonsand their boundaries are based on those in lower levels, and theclassification of high-level polygons can be based upon lower-levelclassifications.The sensor in the aerial camera sees the canopy of the tree, andonce the pine trees are separated from the background (classified),stocking levels can be approximated if the canopy is not completelyclosed. If the canopy is completely closed, it can be difficult to distinguishindividual trees from tree clusters. However, even whenthe canopy is closed, it may be possible to distinguish stockinglevels based on other spectral properties. Therefore, during thedevelopment phase of the project, stocking levels are defined ingeneral terms, such as low density, medium-low, up to high-density.Medium-high and high-density classes are further subdividedinto young trees and old trees, based on other spectral properties(band 3/band 2 ratio). Seven classes are defined initially, with thegoal of aggregating them into two to three meaningful classes toproduce the final map.The choice of the imagery bands, or imagery derivatives touse in the software, is also very important. eCognition internallycalculates many different statistics for each segment on each level,only a few of which are useful for differentiating Lodgepole Pinefrom the surrounding land cover. To winnow these choices downto a statistically significant few, the segment statistics (independentvariables) were input to a statistical modeling softwarealong with a training sample (Pine and Non-Pine, the dependentvariable). The software chose image ratio bands, band 3/band 1,and band 3/(band 1 + band 2 + band 3), as the most likely to discriminateLodgepole Pine from its surroundings. This result madeintuitive sense for two reasons: band 3, or the near-infrared band,is more sensitive to chlorophyll content than the visible-lightbands, and therefore better for discriminating different types ofvegetation. Also it makes sense because band ratios, for exampleband3/band1, tend to normalize the imagery for lighting and terraineffects. An added virtue of using band ratios is that the valuesare more consistent from image to image, making a rule-basemore ‘portable’, and making automation much simpler.Once the higher-level image classification is complete,showing the desired stocking classes and stand boundaries,it is exported as a shapefile, which may be viewed in many differentsoftware packages (Figure 5). At this point, the standboundaries appear convoluted, with too many extraneous bendsand crenulations. This is because the lines were generated fromthe raster image. To make the boundaries more readable to theforester, the lines can be generalized, or straightened to a degree,while removing unnecessary bends (Figure 6). The degree of generalizationmay be adjusted to provide optimum readability andarea estimates. In this representation, the forester is observinga minimum mapping unit of approximately 5 acres. (No stands24 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


2Level 1 classification of pines using small segments.Pines are green, non-pine is gray3less than 5 acres were delineated.) Errors in the boundaries areevident; these will lead to errors in volume and density estimates.The protocols developed in this project will be transferable toother imagery types, such as scanned aerial photographs, withappropriate modifications to the classification algorithm. The useof near-infrared (NIR, red, green) imagery is recommended dueto the utility of the NIR band for discriminating target species.Once the algorithm is calibrated, creation of the stand boundary/stocking layer from raw image to smoothed polygons requires approximately10-12 minutes for a 2000 x 3600 pixel image on aworkstation with a 2.4Ghz processor with 2Gb RAM.Results of the project will be visually evaluated by Plum CreekGIS Analysts and foresters. Stocking classes will additionally beevaluated with respect to the cruise data currently in Plum Creek’sdatabases, and to further refine the class definitions. It is likelythat some manual processing of the maps will be necessary to optimizethe stand boundaries, but preliminary results indicate thatautomated processing of aerial imagery could provide significantsavings in time and resources for Plum Creek. «6Level 2 classification of stocking levels based on relative area ofpines in each image segment4 Not coniferous forestLow-density coniferous forestMed-low density coniferous forestMedium-density coniferous forestMed-high younger coniferous forestMed-high older coniferous forestHigh-density older coniferous forestHigh-density younger coniferous forestStand boundary polygons after the smoothing algorithm is runLegend for stocking level classification57Unsmoothed stand boundaries derived from the stockinglevel classificationHand-drawn boundaries over the same standw w w . i m a g i n g n o t e s . c o mW I N T E R 2 0 0 425


The Indian Remote SensingProgram (IRS) has emerged as one of the highestprofile programs in the commercial imagingindustry. The IRS Program is the world’slargest with six satellites currently in orbit andoperational, including IRS-1C, P2, P3, 1D, P4and RESOURCESAT-1. Two additional IRSsatellites, Cartosat-1 and RESOURCESAT-2,will be launched over the next two years, andfour additional “next generation” satelliteswill be launched in the future.On Oct. 17, 2003, the Indian remotesensing satellite RESOURCESAT-1 waslaunched into an 817 km sun-synchronouspolar orbit by the Indian Space Research Organization(ISRO). RESOURCESAT-1 is themost advanced satellite built by ISRO, bringingcontinuity to the current IRS 1-C and1-D programs. RESOURCESAT-1 has threesensors providing 5.8-, 23.5-, and 56-meterresolution panchromatic and multispectraldata for a vast variety of applications.Since 1994, Space Imaging has partneredwith Antrix Corporation, a divisionof the Indian Space Research Organization(ISRO), to exclusively market and distributeIRS products and ground stations outside ofIndia. Space Imaging and Antrix have recentlyBy Dr. R.S. RaoDirector, IRS Program, Space Imaging,Thornton Colo.Andrea M. CookManager, International Business Development,Space Imaging, Thornton Colo.New eye in the sky:RESOURCESAT-1concluded agreements to extend this partnershipthrough 2010 and are working to commercializeIRS imagery products and groundstations both inside and outside of India.Space Imaging (Thornton, Colorado)launched IKONOS, the world’s first 1-meter,high-resolution commercial remote sensing satellitein September of 1999. Space Imaging hasdeveloped an extensive network of governmentand commercial Regional Ground Stationsworldwide that have the capability to access,download, process and distribute IKONOSdata and products. Through the partnershipwith Antrix and the Indian Government, SpaceImaging has also established more than 14International IRS Ground Stations worldwide,and many stations will be upgrading their systemsto download RESOURCESAT-1 throughupgrades or multi-source ground stations.RESOURCESAT-1 is designed to supportapplications such as urban planning, nationalsecurity, mapping, agriculture and crop monitoring,forestry, and disaster management.With three sensors on board, RESOURC-ESAT-1 is capable of collecting panchromaticand multispectral data ranging from 5 to 56meters in resolution, with a revisit time of fiveto 24 days.RESOURCESAT-1, the tenth satellite ofISRO in the IRS series, is the most advancedremote sensing satellite built by ISRO.RESOURCESAT-1 is intended not only tocontinue the remote sensing data servicesprovided by IRS-1C and IRS-1D, but alsoto vastly enhance data quality. RESOURC-ESAT-1 carries three sensors similar to thoseof IRS-1C and IRS-1D, but with vastly improvedspatial resolutions.w w w . i m a g i n g n o t e s . c o mW I N T E R 2 0 0 429


HIGH-RESOLUTION LINE AR IM AGING MEDIUM-RESOLUTION LISS IIISELF - SC ANNER (LISS -IV)An improved version LISS-3 camera withLISS-IV operates in three spectral bands four bands (red, green, near-IR and SWIR),in the Visible and Near-Infrared (VNIr) or all at 23.5-meter resolution and 141 kmPAN mode with 5.8 meter spatial resolution; swath, will provide essential continuity toit is steerable up to +/-26 degrees across IRS 1-D data. The sensors on board the satellitewill provide data useful for vegetation-track to obtain stereoscopic imagery andfive-day revisit capability.related applications and will allow multipleLISS-IV, with 5-meter resolution imagery, crop and species discrimination.will be very useful for mapping and creating LISS-III with 23-meter resolution imageryhas proved its utility beyond doubt forspatial databases, large scale maps, urbanplanning and development, and National land use, landcover mapping, agricultureand Homeland Security applications. Remote applications such as crop acreage estimationand yield predicition, damage assess-sensing is used in the agriculture industry toanalyze crop growth and soil characteristics, to ment for disaster management and insurance,forest mapping and management, anddetect change, to predict yield, and for modelsand precision farming. Agriculture applicationscan require frequent revisits of the samea variety of environmental applications.area, and LISS-IV’s 5-meter resolution andday revisit time offers a cost-effective solution.RESOURCESAT-1 Technical SpecificationsOrbitCircular Polar Sun SynchronousOrbit height817 kmOrbit inclination 98.7°Orbit period101.35 minutesNumber of orbits per day 14Local time of equator crossing10:30 amRepetivity (LISS-3)24 daysRevisit (LISS-4)5 daysLift-off mass1360 kgMission life5 yearsADVANCED WIDE FIELD SENSOR (AWiFS)The Advanced Wide Field Sensor (AWiFS)has a 60-meter average resolution with a700 km swath and a five-day revisit. Thiscamera will greatly aid crop and vegetationmonitoring, country-level crop estimations,and water resource applications. This is animprovement to IRS 1-C and 1-D’s AWiFSsensor, which has 180-meter resolution.AWiFS, with twin cameras, offers excellentuse for national crop monitoring for foodsecurity assessment, regional environmentalmonitoring, and disaster management applicationslike flood monitoring, forest firemonitoring, damage assessment, and droughtassessment through temporal Normalized DifferentialVegetation Index (NDVI) changes.Space Imaging has the exclusive worldwiderights to distribute RESOURCESAT-1imagery products outside of India. SpaceImaging offers imagery products anywherein the world, except India, using its receptionstation in Norman, Okla., other IRSInternational Ground Stations, or by using120 Gigabits of on-board memory from theRESOURCESAT-1 satellite. «Payloads LISS-4 LISS-3 AWiFSSpatial Resolution 5.8 23.5 60Swath (km) 23.9 (MX mode) 141 740Spectral Bands(micron)70.3 (PAN mode)0.52 - 0.59 0.52 - 0.59 0.52-0.590.62 - 0.68 0.62 - 0.68 0.62 - 0.680.77 - 0.86 0.77 - 0.86 0.77 - 0.861.55 - 1.70 1.55 - 1.70Quantisation (bits) 7 7 10Data Rate (MBPS) 105 52.5 52.530 W I N T E R 2 0 0 4 w w w . i m a g i n g n o t e s . c o m


2004 events calendarMunichfebruary9 — 10ILMF 2004 Annual MeetingRosen PlazaOrlando, Fla.http://www.lidarmap.org/25 — 26AFCEA HomelandSecurity ConferenceRonald Reagan InternationalTrade CenterWashington, D.C.http://www.afcea.org/homeland04/25 — 27IT/GIS in Public Works ConferenceWestin CharlotteCharlotte, N.C.http://www.urisa.org/PublicWorks/publicworks.htmmarch4 – 53rd International eCognition UserMeetingForum HotelMunich, Germanyhttp://www.definiens-imaging.com/um2004/6 — 9ESRI Worldwide Business PartnerConferenceWyndham Hotel and Palm SpringsConvention CenterPalm Springs, Calif.http://www.esri.com/events/bpc2004/index.html16GeoNorthNew Century HouseManchester, Englandhttp://www.geo-north.com/22 — 26TUGIS 2004, the 17th Annual GISConferenceTowson UniversityTowson, Md.http://cgis.towson.edu/tugis2004/28 — 31GeoTec Event —Pathways to IntegrationMetro Toronto Convention CentreTorontohttp://www.geoplace.com/gt/htm/default.aspapril14 — 1614th Annual Nevada GISState ConferenceOrleans Hotel & CasinoLas Vegas, Nev.http://www.ngis.org/18 — 222004 MidAmerica GIS SymposiumHyatt Regency Crown CenterKansas City, Mo.http://magicweb.kgs.ku.edu/19 — 232004 Intermountain GIS Conference— Tools of DiscoveryBillings, Mont.http://www.intermountaingis.org/20-22ITEC 2004ExCel CentreLondonhttp://www.itec.co.uk25 — 28GITA 27th Annual ConferenceWashington StateTrade & Convention CenterSeattle, Wash.http://www.gita.org/events/events2.htmlToronto28 — 31Integrating GIS & CAMA ConferenceHiltonAustin, Texashttp://www.urisa.org/cama.htm28 — 31GIS-T: The Many Faces of GISBest Western Ramkota HotelRapid City, S.D.http://www.gis-t.org/31 — April 2Annual Texas GIS ForumJ.J. Pickle Research CenterAustin, Texaswww.tnris.state.tx.usw w w . i m a g i n g n o t e s . c o mW I N T E R 2 0 0 431


12076 Grant StreetThornton CO 80241 USA

More magazines by this user
Similar magazines