society of american foresters western forester · 2013. 1. 24. · scape at unprece-dented...

of 13/13
BY STEVE REUTEBUCH AND BOB McGAUGHEY Over the last decade, a revolution in active remote sens- ing technology has occurred, providing new tools for measur- ing and monitoring forests over the land- scape at unprece- dented resolution and accuracy. The basis of this revolu- tion is the ability to directly measure the three-dimensional structure (i.e., terrain, vegetation and infra- structure) of forests and to separate measurements of above-ground vegetation from meas- urements of the terrain surface. Of these new remote-sensing technolo- gies, airborne laser scanning, a type of light detection and ranging (LIDAR), is the most commonly available (see sidebar: Airborne LIDAR in a Nutshell). Nationally, at least 30 remote sens- ing companies have LIDAR sensors and are providing LIDAR for a wide range of applications. Several eastern states have embarked on, or completed statewide LIDAR acquisitions primarily for natural hazards mapping, particu- larly updating of flood zone maps. In Oregon and Washington, two public LIDAR acquisition consortiums were formed, initially focused on the heavily forested areas west of the Cascades (see sidebar: How to Get LIDAR Data). As in other states, map- ping of natural hazards (earthquake faults in the Puget Sound trough and landslides in western Oregon) has been the main justification for these efforts. However, participating consor- tium partners have recognized and are encouraging the use of these publicly available LIDAR datasets for other uses, particularly forest management. Much research is underway to develop more precise measures from LIDAR; however, the following simple LIDAR-derived products are easily generated and quite useful to resource managers. High-resolution ground surface models. Traditional digital terrain models (DTMs) were compiled from aerial photos that required map mak- ers to make their best guess about where the ground surface was in heavily forested areas. LIDAR can provide much more accurate ground models for slope mapping, stream delineation, and road and harvest system planning and design. The Oregon and Puget Sound LIDAR Consortiums are producing DTMs with one- to two-meter grid resolu- tion, a vast improvement over the standard USGS 10-meter DTMs. Canopy height models. By sub- tracting the LIDAR-derived ground surface DTM from a LIDAR-derived canopy surface model, a canopy height model (CHM) is produced. CHMs provide spatially-explicit stand structure data over the landscape for estimation of growing stock, input for habitat and fire models, and any other resource planning activities where spatial arrangement and tree height are important considerations. Percent canopy cover models. These models provide a direct meas- urement of cover by height above ground. LIDAR intensity images. These high-resolution images can be matched with existing orthopho- tographs and other digital imagery for change detection and monitoring over time. Intensity data from leaf-off acquisitions can be used to separate hardwood from conifer canopy areas; intensity data from leaf-on data can be used to separate live trees from dead trees. All Returns Datasets. This archive of the LIDAR point cloud (including all returns for each pulse) provides baseline data on current terrain and vegetation structure that is valuable for future change detection and mon- itoring (e.g., crown expansion or dieback). These files can also be used when checking the quality of other derived LIDAR products. For instance, the point cloud can be superimposed on the LIDAR ground surface model to assess how well the ground fits the raw LIDAR scan. Most LIDAR vendors can easily pro- vide these simple products along with the raw LIDAR point data. Also, public domain software is available from the U.S. Forest Service that can be used to process, visualize and perform basic measurements with LIDAR data (see In This Issue: LIDAR LIDAR:An Emerging Tool for Multiple Resource Measurement, Planning and Monitoring (CONTINUED ON PAGE 2) SOCIETY OF AMERICAN FORESTERS March/April 2008 Oregon • Washington State • Inland Empire • Alaska Societies Volume 53 • Number 2 Western Forester Steve Reutebuch Bob McGaughey

Post on 20-Aug-2020




0 download

Embed Size (px)



    Over the lastdecade, a revolutionin active remote sens-ing technology hasoccurred, providingnew tools for measur-ing and monitoringforests over the land-scape at unprece-dented resolutionand accuracy. Thebasis of this revolu-tion is the ability todirectly measure thethree-dimensionalstructure (i.e., terrain,vegetation and infra-structure) of forestsand to separate measurements ofabove-ground vegetation from meas-urements of the terrain surface. Ofthese new remote-sensing technolo-gies, airborne laser scanning, a type oflight detection and ranging (LIDAR), isthe most commonly available (seesidebar: Airborne LIDAR in a Nutshell).

    Nationally, at least 30 remote sens-ing companies have LIDAR sensorsand are providing LIDAR for a widerange of applications. Several easternstates have embarked on, or completedstatewide LIDAR acquisitions primarilyfor natural hazards mapping, particu-larly updating of flood zone maps.

    In Oregon and Washington, twopublic LIDAR acquisition consortiumswere formed, initially focused on theheavily forested areas west of the

    Cascades (see sidebar: How to GetLIDAR Data). As in other states, map-ping of natural hazards (earthquakefaults in the Puget Sound trough andlandslides in western Oregon) hasbeen the main justification for theseefforts. However, participating consor-tium partners have recognized and areencouraging the use of these publiclyavailable LIDAR datasets for otheruses, particularly forest management.

    Much research is underway todevelop more precise measures fromLIDAR; however, the following simpleLIDAR-derived products are easilygenerated and quite useful toresource managers.

    High-resolution ground surfacemodels. Traditional digital terrainmodels (DTMs) were compiled fromaerial photos that required map mak-ers to make their best guess aboutwhere the ground surface was inheavily forested areas. LIDAR canprovide much more accurate groundmodels for slope mapping, streamdelineation, and road and harvestsystem planning and design. TheOregon and Puget Sound LIDARConsortiums are producing DTMswith one- to two-meter grid resolu-tion, a vast improvement over thestandard USGS 10-meter DTMs.

    Canopy height models. By sub-tracting the LIDAR-derived groundsurface DTM from a LIDAR-derivedcanopy surface model, a canopyheight model (CHM) is produced.CHMs provide spatially-explicit standstructure data over the landscape forestimation of growing stock, input for

    habitat and fire models, and anyother resource planning activitieswhere spatial arrangement and treeheight are important considerations.

    Percent canopy cover models.These models provide a direct meas-urement of cover by height aboveground.

    LIDAR intensity images. Thesehigh-resolution images can bematched with existing orthopho-tographs and other digital imagery forchange detection and monitoringover time. Intensity data from leaf-offacquisitions can be used to separatehardwood from conifer canopy areas;intensity data from leaf-on data canbe used to separate live trees fromdead trees.

    All Returns Datasets. This archiveof the LIDAR point cloud (includingall returns for each pulse) providesbaseline data on current terrain andvegetation structure that is valuablefor future change detection and mon-itoring (e.g., crown expansion ordieback). These files can also be usedwhen checking the quality of otherderived LIDAR products. Forinstance, the point cloud can besuperimposed on the LIDAR groundsurface model to assess how well theground fits the raw LIDAR scan.

    Most LIDAR vendors can easily pro-vide these simple products along withthe raw LIDAR point data. Also, publicdomain software is available from theU.S. Forest Service that can be used toprocess, visualize and perform basicmeasurements with LIDAR data (see

    In This Issue: LIDAR

    LIDAR: An Emerging Tool for Multiple ResourceMeasurement, Planning and Monitoring


    S O C I E T Y O F A M E R I C A N F O R E S T E R S

    March/April 2008 Oregon • Washington State • Inland Empire • Alaska Societies Volume 53 • Number 2

    Western Forester

    Steve Reutebuch

    Bob McGaughey

  • sidebar: Fusion LIDAR Software).Over the last decade numerous proj-

    ects have demonstrated that LIDARdata can provide high-resolution, spa-tially-explicit information for multi-resource management and planning.Simultaneously, LIDAR has emerged asthe leading technology for high-resolu-tion terrain mapping needed to betteridentify natural hazards such as flood-and landslide-prone areas. As LIDARsensors and vendor capabilities contin-ue to grow, LIDAR data will become asindispensable to tomorrow’s forestersas the aerial photograph has been totoday’s foresters! ◆

    Steve Reutebuch and Bob McGaugheyare research foresters for the PacificNorthwest Research Station, ResourceManagement and ProductivityProgram, USDA Forest Service, inSeattle, Wash. Steve can be reached at206-543-4710 or [email protected] can be reached at 206-543-4713 [email protected]


    Next Issue: Tribal Forestry


    Western ForesterSociety of American Foresters

    4033 S.W. Canyon Rd. • Portland, OR 97221 • 503-224-8046 • FAX [email protected][email protected] •

    Editor: Lori Rasor • Assistant: Michele DocyWestern Forester is published bimonthly by the

    Oregon and Washington State Societies’ Northwest Office

    State Society Chairs

    Oregon: Marc Vomocil, Starker Forests, Inc.,P.O. Box Box 809, Corvallis, OR 97339; 541-929-2477; [email protected]

    Washington State: Zoanne Thomas, P.O.Box 489, Vader, WA 98593; 360-274-2051;[email protected]

    Inland Empire: Jennifer Costich-Thompson,P.O. Box 508, Sagle, ID 86860; 208-255-2056;[email protected]

    Alaska: Jim LaBau, CF, 2951 Admiralty BayDr., Anchorage, AK 99515; 907-344-1018;fax 907-344-0915; [email protected]

    Northwest Council Members

    District I: Kirk David, 24010 N. McCoy Rd.,Athol, ID 83801; 208-683-3168; [email protected]

    District II: Clark Seely, 2600 State St.,Salem, OR 97310; 503-945-7203;[email protected]

    Please send change of address to:Society of American Foresters

    5400 Grosvenor LaneBethesda, MD 20814


    Anyone is at liberty to make fair use of the material in this publication. To reprint or make multiple reproduc-tions, permission must be obtained from the editor. Proper notice of copyright and credit to the WesternForester must appear on all copies made. Permission is granted to quote from the Western Forester if thecustomary acknowledgement accompanies the quote.

    Other than general editing, the articles appearing in this publication have not been peer reviewed for techni-cal accuracy. The individual authors are primarily responsible for the content and opinions expressed herein.

    Golden YearsJohn Grey Wittmeyer was pre-

    sented with his 2007 GoldenAward on January 8 by DarrenMahr. The Coos Chapter mem-ber is a World War II veteran andlong-time employee with theBLM, and still enjoys “carefully”cutting his own firewood.



    The Clackamas County Parks & Forest Department wishes to hire aforester to manage 4,000 acres of county forestland. The position reports tothe Department Manager and includes: timber sale design and on-the-groundpreparation, road and plantation establishment, monitoring and maintenance;timber sale contract administration and accounting; determination of appro-priate measures to protect resources; conducting surveys to determine thebest silvicultural treatments for forest stands; managing public involvementwith timber sale proposals; analyzing natural resource legislation and prepar-ing reports for the Board of County Commissioners; communicating with thepublic regarding tree and forestry related questions, identification and resolu-tion of hazard tree problems in county parks and other county properties.

    A degree in forestry or related field is required. Monthly salary range:$4,200-$5,000 depending on qualifications. Benefits package includes healthand dental insurance, retirement pension, vacation, sick leave and holidays.For more information, contact Chris Van Duzer at 503-353-4663.

  • There are many different types of airborne LIDAR systems, butthe most common for terrain mapping is discrete-return, small-footprint LIDAR. These laser-scanning systems have four majorhardware components: 1) a laser emitter-receiver scanning unit;2) GPS [aircraft and ground units]; 3) a highly sensitive inertialmeasurement unit (IMU) attached to the scanning unit; and 4) acomputer to control the system and store data from the first threecomponents. Large areas are surveyed with a series of swathsthat often overlap one another by 50 percent or more.

    State-of-the-art LIDAR scanners designed for terrain mappingemit near-infrared laser pulses at a high frequency (typically50,000 to 200,000 per second). For each emitted pulse, mostLIDAR sensors can record one to seven reflections from foliage,branches and sometimes the ground as the pulse passes fromthe top of the canopy down through canopy gaps. Using the dis-tance from the sensor to each reflection or “return,” the GPS air-craft position and the IMU aircraft altitude data, a 3D coordinate iscomputed for each object that reflected a pulse, resulting in a rawLIDAR data cloud. In a properly executed mission, the accuracy ofpoints is typically 15 centimeters vertically and 25-50 centimetershorizontally. The density of points collected varies with missionspecifications. In forested areas, one pulse per square meter hasbeen commonly collected in the past; however, many newer sur-veys have collected five-plus pulses per square meter.

    This data cloud (A) is then processed into different productssuch as canopy surface models (B), ground surface models (C),and canopy height models (D). In addition, a series of LIDARpoint cloud metrics can be computed that have been shown to bestrongly correlated with stand mean height, diameter, basal area,volume, biomass, cover and canopy fuel variables. Most LIDARsystems also record the level of near-infrared energy that wasreflected. These return “intensity” values can be used to createnear-infrared images of the forest and to separate leaf-off hard-woods from conifers or dead from live trees. Because the pointcloud is in real-world coordinates, all LIDAR-derived products canbe imported directly into GIS for use with existing orthophotos andresource data layers such as stand polygons and road coverages.


    Airborne LIDAR in a Nutshell

    Schematic of a typical airborne LIDAR system.

    A: LIDAR Point Cloud (4 points/m2); B: Canopy SurfaceModel; C: Ground Surface Model; D: Canopy HeightModel (created by subtracting C from B)

    Top: Aerial photograph. Bottom: LIDAR near-infraredintensity image. Dark areas are conifers; light grayareas are leaf-off hardwoods and dead trees (androads).


    There are dozens of remote sensing vendors that fly airborne LIDAR systems and can provide a range of LIDAR products withcosts ranging from less than $1 to several hundred dollars per acre, depending on mission requirements and desired products.Costs for large blocks over 10,000 acres are generally $1 to $3 per acre for typical LIDAR deliverables. So, before a landownerrequests a bid, it is important that they understand what products and specifications will work for their project. Additionally, LIDARmissions require mobilization of personnel and aircraft, often from other states. Therefore, it is advantageous to spread this mobiliza-tion cost over a large area to hold down the cost per acre. To do this, landowners may want to see if their lands can be flown as partof a larger, coordinated acquisition.

    In Oregon and Washington, consortiums have formed to coordinate large-area LIDAR mapping projects. Both the Oregon LIDARConsortium ( and the Puget Sound LIDAR Consortium ( maps of completed and planned LIDAR project areas. They invite other federal, state, local and private owners to poolresources for more cost-effective LIDAR acquisitions.

    The Oregon consortium’s initial goal is to collect LIDAR over the nominally inhabited areas in western Oregon, with the ultimategoal of covering the entire state. In 2007, the Oregon consortium was funded by the state legislature to collect LIDAR over 2,000-3,000 square milesin 2008. In addition,the Bureau of LandManagement andtribal partners areflying an additional1,000-1,500 squaremiles in Oregon asconsortium partners.The Puget SoundConsortium has col-lected over 12,000square miles andhopes to eventuallyfly all of westernWashington. Theconsortium has alsoflown smaller areasin easternWashington withstate and federalpartners. LikeOregon, the ultimategoal is to fly theentire state ofWashington.

    If a landownerdecides to contractfor LIDAR datadirectly with a ven-dor, rather than part-nering with one of the regional consortiums, review of theconsortium specifications will help them better understandimportant choices that must be made regarding missionspecifications, required accuracy standards and options fordeliverables.

    Above is a table of some of the major mission variablesand typical specifications that should be considered whencontracting for LIDAR projects.

    As with any remote sensing contract, the purchasershould also address who owns the collected data, for whatpurposes, and for what timeframe after the project is com-pleted. Some vendors retain ownership of the raw data andonly license use of delivered products to purchasers.

    This highlights another advantage of partnering with theconsortiums—all data from their projects are put in the pub-lic domain and are carefully archived. This long-term ware-housing of LIDAR missions will become particularly impor-tant as areas are re-flown over time (e.g., after large floodevents, landslides or wind storms) and earlier LIDAR dataare combined with data from later flights for change detec-tion and monitoring purposes, such as tree growth, mortali-ty and wind-throw.

    How to Get LIDAR Data

    Partial List of LIDAR Project Specifications in Forested Areas*

    Top: Overhead view of LIDAR points flown in 1999. Darkareas are ground.Bottom: LIDAR for same area flown in 2003. Notice thecrown expansion for most trees and the missing tree on theright side of the 2003 image. This tree was blown downbetween the 1999 and 2003 LIDAR flights.

    *Adopted from the Puget Sound LIDAR Consortium. Refer to article by Ralph Haugerud elsewhere in this publication formore information on model specifications.


    The PNW Research Stationhas developed a suite of soft-ware tools called “Fusion” thatcan be used to combine LIDARpoint clouds with existingorthophotos, maps and GIS lay-ers.

    Fusion also includes utilitiesto process point clouds intocanopy and ground surface mod-els, canopy metrics (heights,cover, etc.) that can be importedinto GIS for further analysis.

    The Fusion package, alongwith an online tutorial, manualand sample dataset is availablefrom the USFS Remote SensingApplications Center’s website



    Fusion LIDARSoftware

    Background: LIDAR contour lines super-imposed on an orthophoto. Foreground:Sample (from area shown in the black square) of LIDAR point cloud with individualtree measurement.



    n increasingnumber of ven-

    dors offer LIDAR sur-veying services to aclient communitythat includes local,state, federal andtribal governments,private landowners large and small,engineering and land managementfirms, and a handful of researchers.With time, LIDAR survey data shouldbecome a well-understood commodi-ty. But we are not there yet! Manypurchasers of LIDAR survey data stillfind that, on occasion, they do notreceive a product that meets theirexpectations. To avoid this, it is help-ful to have a specification that com-municates to the vendor what theclient desires, and that if met, guaran-tees that a data set will be fit for useand provides a framework for resolvingdisputes over data quality.

    Recently, Susan Nelson (Bureau ofLand Management), Diana Martinez(Puget Sound Regional Council) and I,with advice from several colleagues,wrote a model specification for LIDARdata to be purchased by public agen-cies in the Pacific Northwest. The com-plete specification is available online athttp://pugetsoundLIDAR.ess. The specificationis based on prior experience with sev-eral vendors and multiple acquisitioncontracts. While it is informed by theexperience some of us have with thePuget Sound LIDARConsortium, it is not basedsolely on this experience.Use of this specificationshould ease data interoper-ability, reduce contractingcosts, and facilitate devel-opment of a shared set oftools for manipulatingLIDAR data.

    The model specificationis designed for the PacificNorthwest. It reflects the

    prevalence of young, angular land-scapes, the regional importance offorests and fish habitat, and the need tointelligently guide ongoing urbaniza-tion. It may, perhaps with adjustments,be useful elsewhere. The specificationreflects our perception of LIDAR tech-nology and market conditions as of2007. It should evolve with increasingexperience and changing technology.We know that in at least one aspect(classification of LIDAR returns) thespecification needs to be improved.

    Writing a LIDAR survey specificationpresents a challenge. A good specifica-tion is such that: (1) conformance tothe specification can be readily evaluat-ed; and (2) if data conform to the speci-fication, the data are assured of beingsuitable for the task at hand. Absolutevertical accuracy, typically the founda-tion of topographic surveys, fails thischallenge on both counts. LIDAR datashould be accurate, complete andusable. We wrote a specification thatdescribes these qualities and for whichconformance can, with a few excep-tions, be easily measured. In general,the specification focuses on LIDARdata, not the procedures employed tocollect the data. An exception is GPSpractice, as we have found that it is veryexpensive to adequately judge the qual-ity of absolute spatial positioning; forthis reason, we specify some aspects ofGPS procedures.

    In addition, the specification pre-scribes some aspects of GPS proce-dures, prescribes a data-tiling schemeand file names, discusses the negotia-tion of point-classification procedures,

    and provides instructions for formalmetadata. Perhaps the most impor-tant feature of the specification is notthe particular set of choices for pointdensity, absolute accuracy, maximumscan angle, swath overlap and the like,but the recognition that these thingsshould be specified.

    Constraints in survey design

    There are tradeoffs between surveydesign, cost, accuracy and resolutionof a LIDAR survey. To a first approxi-mation, cost is the sum of mobiliza-tion expenses (including establishingGPS ground control), aircraft and crewtime, and processing time. Accuracy iscontrolled by GPS base-line length,inertial measurement unit (IMU) qual-ity, care and experience in calibration,and flying height. Resolution is mostlya function of on-ground spot spacing,which is governed by instrument pulserate, flying height and airspeed. Inforested areas, ground resolution issignificantly decreased as most laserpulses do not produce returns fromthe ground surface.

    At a given pulse density, single-swath(no overlap) data generally provide bet-ter relative accuracy, and thus betterfeature recognition, but may require ahigher pulse-rate instrument to achievethe desired pulse density. However,multiple, overlapping swaths make iteasier to achieve high pulse densitiesand generally have multiple look angles,both desirable characteristics forincreasing the probability of groundreturns in dense forest canopy, but at acost of poorer feature recognition

    A Model Specification for LIDAR Surveys inthe Pacific Northwest


    Summary of the specification

  • because of swath-to-swath errors. Leaf-off acquisition gives much betterground penetration, but at the cost of ashorter acquisition season that general-ly has poor weather. Leaf-on acquisi-tion is likely to be cheaper because ofbetter instrument availability and gen-erally better weather.

    Since 2000, there has been a six-foldincrease in instrument pulse rate.Faster computers, better codes andmore experience have allowed han-dling of greater data volumes at thesame cost. Rather than moving towardlower-cost surveys at the same resolu-tion, the Puget Sound LIDAR Consor-tium has chosen to acquire surveyswith a higher pulse density. There areseveral reasons for this.

    First, and best documented, is thathigher-density surveys allow much bet-ter characterization of the forestcanopy. Second, we observe that witha six-fold increase in pulse density, wehave not seen a correspondingincrease in the number of identifiedground points, but we see fewer vege-tation returns misidentified as groundand fewer landscape corners misiden-tified as vegetation. We suspect thatwith a smaller fraction of returns iden-tified as ground, the confidence thatthese are indeed ground pointsincreases and the fraction of errorsdecreases. The resulting bare-earthsurface models are more detailed andmore informative (see Figure 1). Third,closer pulse spacing probably results inbetter survey calibration, as (a) thedata set has better XY resolution (a lim-iting factor in some pointing calibra-tion procedures); and (b) interpolationerrors associated with tying quasi-reg-ularly spaced LIDAR returns to arbi-trary ground control points are smaller.Fourth, reducing cost by increasing theinstrument pulse rate and flying higherand faster significantly reduces surveyaccuracy, as the dominant error inmost LIDAR surveys is mis-positioningbecause of pointing error and thiseffect is linear with instrument height.

    Why not save money bypurchasing lower quality data?

    LIDAR data are expensive and webelieve that community support forcontinued data acquisition is morelikely if we meet the needs of mostpotential users. Angular landforms,

    dense forest cover, and significant landand habitat values dictate that in thewet Pacific Northwest we need dense,accurate surveys.

    A faster instrument allows one to flyhigher and faster and cover the samearea at the same pulse density in lesstime—but in most cases such cheaperdata will be significantly less accurate.For earthquake and landslide hazardsmapping where public safety is anissue, we are concerned to deflect lia-

    bility issues by using the best-availabledata. In the long run, change detectionand analysis is likely to be a major useof LIDAR data and the ability to detectand describe change is closely relatedto data resolution and accuracy. ◆

    Ralph Haugerud is a research geologistwith the U.S. Geological Survey, sta-tioned at the University of Washingtonin Seattle. He can be reached at 206-553-5542 or [email protected]


    7150 S.W. Hampton St., Suite 228Portland, OR 97223-8378

    503-684-5727 •

    Experts in Forest Valuation and Investment Analysis


    Figure 1. Bare-earth images showing effects of pulse density in forestedareas. Top, 2005 survey at ~2 pulses/m2. “Crystal forest” is indicative oftoo-few ground returns. Bottom, 2007 survey of same area at ~8 pulses/m2.Note North-South forest road for scale.


    emote-sensed data collected usingLight Detection and Radar

    (LIDAR), when combined with addi-tional data in a two-stage sampling pro-cedure can provide stand-level invento-ry information with sampling errorsthat are equivalent to ground samplingtechniques. The use of remote-senseddata has the potential to reduce fieldcosts and quickly complete large inven-tory programs. These savings are possi-ble because of the economies of scalethat are achieved when computingcapacity is substituted for labor costs.

    The first stage in the two-stage

    process is to acquire the raw LIDARdata. Organizations such as the PugetSound LIDAR consortium and theOregon LIDAR consortium can assistin this process by coordinating theefforts of multiple landowners toacquire LIDAR data. The savings canbe substantial between large andsmall projects.

    LIDAR acquisition in the 5,700-acrePanther Creek watershed approached$5.00 per acre, whereas the cost of dataacquisition in the 1,500,000-acre CoosBay project coordinated by the OregonLIDAR consortium is expected to beless than $0.70 per acre. In addition tocoordinating large acquisition proj-

    ects, the consortiums can establishtechnical specifications for resource-grade LIDAR acquisitions that willensure the compatibility of currentand future LIDAR collection projects.

    Raw LIDAR data can be used tomodel the surface of the earth andindividual tree canopies (ITC). Theinformation from the bare-earth mod-els can be used to produce severalproducts, such as digital elevationmodels, without the need to acquireadditional data. The ITC models canbe used to locate individual treecanopies and estimate canopy height,area, shape and return intensity, but inorder to produce a stand-level invento-ry, additional information is required.

    The second stage of the data acquisi-tion process involves acquiring infor-mation that can be used to identify treespecies and to estimate individual treediameters and the number of ITC poly-gons that contain more than one tree.The key to creating a stand-level inven-tory from LIDAR data is establishing anunbiased link between remotely sensedITC data and ground measurements ofidentifiable trees. This link enables theintegration of additional remote-senseddata and the establishment of a statisti-cal correlation between the ITC dataand ground measurements.

    Digital color infrared (CIR) photog-raphy is used to identify the species ofindividual trees through a process thatrequires the digital CIR photography tobe intersected with the ITC polygonlayer. This produces a layer where thespectral characteristics of individualtree canopies can be identified andspecies identified based upon the spec-tral characteristics, provided there is agood match between the location of theITC polygons and the CIR photography.

    The correlation process to identifyindividual tree metrics requires theboundaries of individual stands in theinventory area be identified as well asstratifying the stands based upon theirLIDAR derived metrics. A statisticalanalysis is conducted to determine thenumber of plots and stands where cor-relation plots are to be established. It isimportant that the ground measure-ments are accurate and unbiased at thestand level. This requires that the trees

    Developing a Stand-Level Inventory Using LIDAR




    The Brightest New Herbicide forHerbaceous Weed Control

    for Douglas-fir and other Conifers.

    —See your Wilbur-Ellis Pro for more details—

    9685 Ridder Rd. S.W., Suite 190 • Wilsonville, OR 97070

    For information on this or any other herbicides, call:

    Bruce Alber 503-227-3525–western Oregon & western Washington

    Joel Fields 509-928-4512–eastern Washington, Idaho & Montana

    Scott Johnson 916-991-4451 & Jerry Gallagher 530-570-5977–California

  • on the correlation plots be located withsurvey-level accuracy in order toachieve as near to a one-to-one rela-tionship between the ITC location andthe ground-based tree location. Tests ofthis process by ImageTree Corporationin the southeast United States haveyielded root mean square errors forbasal area and volume estimates at thestand level of 9.7 and 12.8 percent.

    Once the correlation is established inthe sample stands, this informationalong with the CIR photography and theITC map is used to develop stand tablesfor all of the stands in the inventoryproject. If there is a good correlationbetween the modeled stand boundariesand the actual stand boundaries, thenthe stand table information derivedfrom the remote sensed data is expectedto be similar in accuracy to ground-based inventories.

    The knowledge base necessary tocomplete stand level inventories usingremote-sensed data is expanding.Several organizations are developingand validating the computer algo-rithms necessary to adapt this sam-

    pling process to the forests of thePacific Northwest.

    Developing a stand-level inventoryusing remote-sensed data is a comput-er-intense process. The raw LIDARdata for the Coos Bay project alone isexpected to approach 2.5 terabytes. Alimited number of organizations havethe knowledge and computer capacitynecessary to work with multi-terabytefiles. The use of a computer-intensiveinventory process to replace a labor-intensive process provides economiesof scale that reduces the cost per acrefor large inventory projects.

    In the 5,700-acre Panther Creek, thecost for a stand-level inventory basedupon remote sensing would approach$20.00 or more per acre. The Washing-ton State DNR is currently completinga 200,000-acre inventory project usingremote-sensed data and the cost isapproximately $4.00 per acre, includ-ing the $1.50 per acre for acquisition ofthe raw LIDAR data.

    The $4.00 per acre cost of remote-sensed inventories compares favorablywith the $7.00 to $10.00 per acre that a

    typical ground-based inventory costs tocomplete. In the future, the cost of aninventory based upon remote-senseddata should decrease as computer pro-cessing and storage become cheaperand the algorithms necessary to pro-duce the inventory become morerefined. The cost of a ground-basedinventory will likely increase in thefuture in relation to labor cost. Theresult is that stand-level inventoriesusing remote-sensed data will becomemore cost competitive in the future.

    The use of LIDAR as part of aninventory program will not completelyreplace the need for ground-basedinventories. Stand cruises will still benecessary in small-scale projects, inhigh-value projects that require lowstandard errors, and in any stand typewere LIDAR cannot produce equiva-lent estimate errors. ◆

    George McFadden is silviculturist withthe Bureau of Land Management-Oregon State Office, in Portland. Hecan be reached at 503-808-6107 [email protected]


    High speed cruising.









    Cruising has never been so fast and easy! The Haglöf Laser Vertex Hypsometer combines a laser, ultrasound and tilt sensor, into a single precision instrument that provides accurate and reliable distance, height, and angle measurements over and over. And the Laser Vertex is just one of the more than 9,000 top-quality products geared specifically to outdoor pros. And each item comes with unsurpassed tech support and customer service, plus our 100% Satisfaction Guarantee! Give us a call or log on to to get a free copy of our latest catalog today!

    Catalog Request: 800-360-7788Sales: 800-647-5368

    ©2008 Forestry Suppliers, Inc. All rights reserved.


    ecent advances in remote sensingand data collections are creating

    an environment where forest opera-tion designs result in a high level ofagreement between paper designs and

    their corresponding field locations. Infact, we see a change in paradigm inthat discrepancies between map-based locations and field-located con-ditions are not the result of poor mapmaterial, but rather a reflection of nolonger appropriate (or needed) meas-uring and referencing procedures dur-ing the field verification process. This

    article discusses the impact of theincreased quality of data collection onroad and skyline profiles.

    Maps have been one of the criticaldata requirements for forest engineer-ing application, ranging from topo-graphic to forest stand maps. In thepast, ground-based and photogram-metric mapping has been the mostcost effective way to build topographicand other maps of forested areas. Atthe turn of the last century, during therailroad logging days, the necessarydetailed maps were created using staffcompass and steel tapes. Those mapsdid have the necessary level of detail,usually with 0.5 meter contour inter-val, unencumbered by tree coverage.In later years, the advent of aerial pho-tography led to the creation of pho-togrammetric maps. These maps pro-vide good preliminary guidance forlaying out roads and harvest units;unfortunately, the trees that draw us tothese areas also obscure the underly-ing topography. In difficult topogra-phy, planned skyline profiles and road


    The Impact of LIDAR Technology on TransportationSystem Design: Moving from Coarse TopographicMaps to Detailed Digital Elevation Models


    Peter Schiess

    Figure 1. LIDAR topography provides detail from road beds, individual slashpiles, ditches and earth slumps. Note the earth slump encroaching on anexisting road with the headwall clearly noticeable. Walking the ground, fieldengineers were not even aware of the headwall.

  • alignments are frequently renderedunworkable by topographic “details”that are not represented in the pho-togrammetric topography that is usedto plan them.

    For that reason, forest engineersalways emphasized the importance offield verification. Initial planning inthe office was certainly recognized asimportant, but its primary functionwas to focus field reconnaissance.Field reconnaissance always has beentime consuming and therefore expen-sive. Due to the often long “walk-in”times to get to the necessary planninglocations, substantial time had to beallowed for, or limited field verificationwas done to stay “on budget.”

    Recent technological advances ledto a rapid spread in airborne laseraltimetry (LIDAR) mapping of theearth’s surface. Just as in photogram-metry, forest canopies can interceptmost of the laser pulses, but any standin which sky can be seen from theground will allow LIDAR penetration tothe ground. Wherever the LIDAR pulsedensity can overcome canopy density,the resulting ground points can beinterpolated into a topographic map.

    The detail of the new LIDAR-gener-ated maps can be seen in the comput-er-generated hill-shaded image (Figure1). In addition to roads and streams,roadside ditches are clearly evident, asis a subtle earth slumping along theeastern edge. The minor mounds scat-tered across the area in the upper halfof Figure 1 appear to correspond withstumps and slash piles. This newmapping shows considerable promisein a range of designs from skyline cor-ridor profiles to road location anddesign activities.

    Recent experiences with the use ofLIDAR-generated maps as part of theUniversity of Washington ForestEngineering (FE) Senior projects, incollaboration with Washington StateDepartment of Natural Resources(DNR) has led to a significant shift inhow to approach paper planning andsubsequent field reconnaissance. Aspart of the planning for the TahomaState Forest, the FE seniors developedLIDAR-based paper plans that weresubsequently field verified. For a par-ticular timber sale DNR had field-measured skyline profiles to assure

    their technical feasibility. Field worktook about a one-person-day given thedifficulty of terrain, brush conditions,etc. The LIDAR maps provided amuch more realistic assessment than

    the photogrammetically-derived mapcould (Figure 2). The LIDAR mapsshowed a much higher level of detailthan the standard maps did. Not onlythat, but the LIDAR profile could be


    Figure 2. Shown are a LIDAR-generated contour map with a two-meter pixelsize (left) and a photogrammetically-derived contour map (right, with a six-meter pixel size) showing a field-verified profile as well as profiles based onthe DNR and LIDAR maps. The LIDAR map clearly identifies a bench (arrow)not shown by the DNR contour map. Also note the topographic detail of theLIDAR map elsewhere, which the DNR map does not display.

    [email protected]

  • generated inless than fiveminutes, com-pared to theone-person-dayto generate thefield profile.The time sav-ings are obvi-ous.

    It is speculat-ed that theLIDAR profileprovides thebest approxima-tion of trueground condi-tions given theprecision of thefield instru-ments used (clinometers, hand com-pass and string box). Other researchersestablished high correlations betweenLIDAR-derived topography and truetopography based on terrestrial map-ping.

    The nature of the map location(pegging) and grade-line locationprocess is now beginning to change aswell. Great emphasis can now be puton pegging roads on LIDAR-derivedmaps. A pegging tool that automatesthis process is available from the RuralTechnology Initiative website (Figure 3, Thepegged roads on digital LIDAR mapsare the script, laid out in the office,and GPS units can be used to keep theroad locators on track in the field (“fol-lowing the script”).

    The purpose now is to “find thelocation on the ground” as predictedby the paper road location, rather thanbeing guided by the simple fieldinstruments such as cloth tape, handcompass and clinometers. Differencesusually are due to the metrics used inthe field for establishing grade lines.Those field instruments are far lesssophisticated and less precise than theprocess of establishing a paper-maproad from a LIDAR DEM (digital eleva-tion model). The road locationprocess is now moving from a fieldverification process to a “field trackingof map-derived (LIDAR DEMs) roadlocations,” a basic change in road-location paradigm. Office-locatedroads can now be exported intoRoadEng (a commercial forest road

    design package) for further evaluationof critical areas such as switchbacks orstream crossings (Figure 4).

    LIDAR-derived maps with their high

    resolution also offer new ways to lookat terrain features. Traditional mapsutilize contour lines of varying equidis-tance, typically 20 feet for maps of


    Figure 4. Road systems developed with PEGGER based on two-meter DEMs.On the left, the dashed line represents the pegged road. The circle with dotrepresent GPS location points collected in the field during the field reconnais-sance phase. The triangles are proposed landings. The underlying DEM has agrid spacing of two meters. The enlarged area shows a plan and profile viewof a critical switchback road location, initially pegged with PEGGER and fur-ther evaluated by exporting the data into RoadEng for additional evaluation,work all done in the office.

    Figure 3. Locating a road with PEGGER. The tool allows for rapid road location on a digital map.LIDAR-derived maps now have such precision that map-located roads agree very well with subse-quent field verification to the point that those map locations can almost be accepted as “field-veri-fied.” However, issues such as seepage and rock outcrops do not yet show up on these LIDAR maps,so some field verifications are still needed.

  • 1:4800 scale ratios. Students com-mented that the slope class maps pro-vided a much better assessment oflocating oneself in the field as well asfinding critical topographic featuressuch as benches, appropriate areas forswitchback locations and more withLIDAR-derived maps than traditionalcontour maps could provide (Figure 5).

    A question that still has to beanswered conclusively is: Are roaddesign data derived from LIDAR mapsas reliable as road data customarilytraversed in the field? Many have sug-gested that a LIDAR map cell size ofbetween 1.0 and 3.0 meters is suffi-cient for operational route location.Extracting road design data fromLIDAR maps appears feasible whenconsidering that for forest roads, con-struction tolerances quite often are inthe same range.

    Cross section data collected from atraverse were compared with crosssections derived from LIDAR DEMscollected with typical forestry surveyequipment (Figure 6). The traversefield data collected with staff compassand Laser Impulse instrument for dis-tances and slopes were superimposedover the LIDAR DEM and the corre-sponding cross section extracted andcompared with the cross section datacollected in the field (Figure 6).

    From experience, the author willaccept the cross section data fromLIDAR DEMs as comparable to fieldmeasurements where side shots are

    customarily collected with hand-heldclinometers and cloth tape. In fact, itappears that LIDAR-derived cross sec-tion data are of equal if not betterquality, and may provide more detail

    than typically recorded in the field. It appears that we can indeed carry

    out a full road design based on LIDARtopography without going to the field.The issue now becomes one of beingable to find and stake the coordinatevalue of the centerline on the ground.Currently, investigations are on-goingto answer this question. TypicalGlobal Positioning Systems (GPS) unitshave been unreliable, with an absoluteerror in the one- to five-meter rangeunder dense canopy. So our gains ofbetter topographic data are currentlynegated by our inability to accuratelylocate a map-derived point (coordi-nate) on the ground. We currently aretesting some new technologies thathopefully will overcome this hurdle. ◆

    Peter Schiess is a professor of ForestEngineering at the College of ForestResources, University of Washington inSeattle. He can be reached at 206-543-1583 or [email protected]


    Figure 5. Slope class map derived from LIDAR DEM with two-meter grid spac-ing. Slope classes are in varying shades of green, the darker the steeper.Slope class depiction provides much more detail about critical terrain fea-tures than contour lines would. For example, small areas of gentle terrain(colored white or light green) within larger areas of steeper slope classes(dark green) are clearly shown. Note old skid trail locations in the lower rightand roads in the left upper half.

    Figure 6. Road traverse superimposed on LIDAR DEM with cross section loca-tion shown. The small dots are GPS coordinates to geo-reference the fieldtraverse. Both sets identify the slope break below the center-line stake.However, the field crew only took one “side shot” or measurement upslopefrom the traverse point. The LIDAR-derived cross section reveals a breakabove the traverse point.