Light Detection and Ranging (LIDAR): An Emerging Tool for ...

Light Detection and Ranging(LIDAR): An Emerging Tool forMultiple Resource Inventory

Stephen E. Reutebuch, Hans-Erik Andersen, andRobert J. McGaughey

Airborne laser scanning of forests has been shown to provide accurate terrain models and, at the sametime, estimates of multiple resource inventory variables through active sensing of three-dimensional(3D) forest vegetation. Brief overviews of airborne laser scanning technology [often referred to as “lightdetection and ranging” (LIDAR)] and research findings on its use in forest measurement and monitoringare presented. Currently, many airborne laser scanning missions are flown with specifications designedfor terrain mapping, often resulting in data sets that do not contain key information needed forvegetation measurement. Therefore, standards and specifications for airborne laser scanning missionsare needed to insure their usefulness for vegetation measurement and monitoring, rather than simplyterrain mapping (e.g., delivery of all return data with reflection intensity). Five simple, easilyunderstood LIDAR-derived forest data products are identified that would help insure that forestry needsare considered when multiresource LIDAR missions are flown. Once standards are developed, there isan opportunity to maximize the value of permanent ground plot remeasurements by also collectingairborne laser data over a limited number of plots each year.

Keywords: LIDAR, airborne laser scanning, forest inventory, forest structure, forest monitoring

T he goal of forest inventory is to pro-vide accurate estimates of forestvegetation characteristics, includ-

ing quantity, quality, extent, health, andcomposition within the area of interest. Aforest inventory is an estimate of the makeupof plants (primarily trees) that compriseaboveground forest biomass. Ideally, a forestinventory system should be designed to pro-vide spatial data that can be used over a rangeof scales to support a wide variety of resourcemanagement goals for a particular forest, in-cluding silviculture, harvest planning, habi-tat monitoring, watershed protection, andfuel management. However, traditionalground-based forest inventory methods aredesigned to provide point estimates of in-ventory parameters for relatively large areas

to a desired level of precision and are notdesigned to provide spatially explicit, high-resolution mapped information regardingthe spatial arrangement or structure of forestbiological components over the landscape(Schreuder et al. 1993). However, such“biospatial” data are important in all aspectsof natural resource management: the“where” often is as important as the “what.”For most resource management activities,these biospatial data, characterizing how for-est structure and composition vary over thelandscape, are at least as important in eco-nomic, aesthetic, and habitat assessments asare geospatial data (e.g., slope, aspect, andelevation).

Over the last 10 years, a revolution inremote sensing technology has occurred,

providing new tools for measuring andmonitoring biospatial data across thelandscape. The basis of this revolution isthe ability to measure directly the three-dimensional (3D) structure (i.e., terrain,vegetation, and infrastructure) of imagedareas and to separate biospatial data (mea-surements of aboveground vegetation)from geospatial data (measurements of theterrain surface) using active remote sens-ing technologies. Active sensors emit en-ergy (e.g., light or radio waves) and recordthe reflection of this energy down throughthe depth of the canopy. Two active re-mote sensing systems currently are com-mercially available with this capability: (1)airborne laser scanning, also referred to aslight detection and ranging (LIDAR), and(2) interferometric synthetic aperture ra-dar (IFSAR; also referred to as InSAR). Ofthese systems, LIDAR is more technicallymature and widely available, although IF-SAR holds much potential for landscape-level applications. In this article we fo-cused on LIDAR as a tool for multipleresource inventory.

Brief Overview of AirborneLIDAR Technology

There are several varieties of airborneLIDAR systems; in this article we focused onthe most common terrain mapping system,namely, discrete-return, small-footprintLIDAR (i.e., typical laser beam diameter atground level in the range of 0.2–1.0 m). Dis-crete-return airborne LIDAR systems were

286 Journal of Forestry • September 2005

AB

ST

RA

CT

developed over the last 15 years for the ex-press purpose of mapping terrain (Wehr andLohr 1999). Airborne laser scanning systemshave four major hardware components: (1) alaser emitter-receiver scanning unit, (2) dif-ferential global positioning systems (GPS;aircraft and ground units), (3) a highly sen-sitive inertial measurement unit (IMU) at-tached to the scanning unit, and, of course,(4) a computer to control the system andstore data from the first three components.

Laser scanners designed for terrainmapping emit near-infrared laser pulses at ahigh rate (typically 10,000–100,000/sec-ond). The precise position and attitude ofthe laser scanner unit at the time each pulseis emitted are determined from flight datacollected by the GPS and IMU units. Therange or distance between the scanner andan object that reflects the pulse is computedusing the time it takes for the pulse to com-plete the return trip distance from scanner toobject. This range information and the po-sition and orientation of the scanner are usedto calculate a precise coordinate for each re-flection point.

A swath of terrain under the aircraft issurveyed through the lateral deflection of thelaser pulses and the forward movement ofthe aircraft. The scanning pattern within theswath is established by an oscillating mirroror rotating prism, which causes the pulses tosweep across the landscape in a consistentpattern below the aircraft (Figure 1). Largeareas are surveyed with a series of swaths thatoften overlap one another by 20% or more.This results in acquisition of a 3D “pointcloud” from vegetation and terrain, oftenwith several million measurements persquare kilometer. The final pattern of pulsereflection points on the ground and the

scanned swath width depend on the settingsand design of the scanning mechanism (e.g.,pulse rate, returns per pulse, and scanningangle), as well as other factors such as flyingheight, aircraft speed, and the shape of thetopography.

Most LIDAR systems can detect severalreflections or “returns” from a single laserpulse. Multiple returns occur when the pulsestrikes a target that does not completelyblock the path of the pulse and the remain-ing portion of the pulse continues on to alower object. This situation frequently oc-curs in forest canopies that have small gapsbetween branches and foliage. To take ad-vantage of this, most terrain mapping mis-sions over hardwood or mixed conifer-hard-wood forest are flown in leaf-off conditionsto maximize the percentage of pulses thatreach the ground surface. In contrast, whenthe primary objective is characterization ofcanopy conditions, LIDAR missions aresometimes flown in leaf-on conditions tomaximize the number of returns from treecrowns and other vegetation layers.

System manufacturers have expendedgreat efforts to develop methods for distin-guishing between laser reflections from theground surface (terrain measurements) andthose from vegetation. LIDAR system man-ufacturers typically quote root mean squarederrors of 10–15 cm vertical and 50–100 cmhorizontal for terrain mapping products un-der optimal conditions. In several studies thevertical accuracy of LIDAR terrain measure-ments was found to be in the range of 15–50cm over a variety of ground and cover con-ditions from open flat areas (Pereira andJanssen 1999) to variable forest cover (rang-ing from clearcuts to mature stands; Krausand Pfeifer 1998, Reutebuch et al. 2003).

Airborne LIDAR scanning system ca-pabilities have dramatically increased overthe last 10 years. Data acquisition costs havecorrespondingly decreased as advances in in-ertial navigation systems, computing capa-bility, and GPS technology have allowedLIDAR to move into the mainstream com-mercial terrain mapping sector. Today, sev-eral vendors market LIDAR systems, andseveral third-party vendors offer specializedLIDAR data processing software for efficientterrain mapping. Numerous LIDAR surveyfirms offer a complete range of mapping ser-vices including the generation of digital ter-rain models, contour maps, extraction of in-frastructure locations and characteristics,and delivery of processed scanner data in avariety of formats.

LIDAR-Derived ForestMeasurements

Although the mapping community hasembraced LIDAR as the standard technol-ogy for collecting high-resolution geospatialdata over vegetated areas, the natural re-source management community has beenslower to appreciate the capability of LIDARto simultaneously collect high-resolutionbiospatial data. System manufacturers havelargely ignored the potential uses of theLIDAR vegetation returns (that are under-standably considered “noise” in the contextof terrain mapping), and only in the last fewyears have natural resource scientists begunto realize the accuracy and value of LIDARbiospatial forest structure data, with Canadaand Europe at the forefront (Wulder et al.2003, Olsson and Næsset 2004). There are amultitude of uses for such 3D forest struc-ture data, not the least of which is forest in-ventory and monitoring. Several Europeancountries have initiated programs to useLIDAR for large-scale forest inventory;however, forest analysis procedures are notas well refined as are those for terrain map-ping products. Scandinavian researchers(Næsett et al. 2004) have reported generallyvery good results with LIDAR measure-ments of height, volume, stocking, and basalarea in coniferous areas with LIDAR pointdensities ranging from 0.1 to 10 points m�2.Although there is growing interest in the op-erational use of LIDAR for large-scale re-source inventory applications in the UnitedStates, to date, most of the activity has beenlimited to research applications.

LIDAR-Based Measurement ofIndividual Tree Attributes

Individual tree crowns composing thecanopy surface can be detected and measuredautomatically with relatively high accuracythrough the application of computer vision al-gorithms (Figure 2) when LIDAR data are ac-quired at a high density (4–5 points m2). Sev-eral studies have shown that when the canopyis composed of a single canopy stratum, mor-phological computer vision techniques can beused effectively to identify automatically treecrown structures and measure individual treeattributes, including total height, crownheight, and crown diameter (Ziegler et al.2000, Persson et al. 2002, Schardt et al. 2002,Andersen 2003, Straub 2003). Popescu et al.(2003) have shown that although individualtree heights can be estimated using lower-den-sity LIDAR data (1 point m�2), it is difficult to

Figure 1. Schematic showing LIDAR datacollection over bare ground.

Journal of Forestry • September 2005 287

measure accurately other crown attributes,such as crown width, especially in mixed de-ciduous forest types. Several studies have

shown that the combined use of LIDAR andmultispectral digital imagery can lead to moreaccurate individual tree- and plot-level esti-

mates of critical inventory variables such asheight, stem volume, basal area, biomass, andstem density (McCombs et al. 2003, Popescuet al. 2004).

When high-density LIDAR data setsare available from different years, the differ-ence in the individual tree canopy measure-ments generated from the multitemporalLIDAR data sets represents an estimate ofthe tree growth over the intervening period(Yu et al. 2004). In Figure 3, one can clearlysee the expansion of individual tree crownsand the removal (due to windthrow in thiscase) of an individual crown by comparingLIDAR data clouds from 1999 and 2003 forthe same strip of forest. The use of multi-temporal LIDAR therefore has the potentialfor monitoring growth and mortality for alloverstory trees within a certain area. As anexample, in a study performed at the CapitolState Forest study area in western Washing-ton State, Andersen et al. (2005a) used high-density LIDAR data acquired in early 1999and late 2003 to extract individual treeheight growth measurements for 1.2 km2 ofmountainous second-growth, naturally re-generated Douglas-fir forest (Pseudotsugamenziesii (Mirb.) Franco var. menziesii).Preliminary results of this analysis showedthat subtle differences in growth betweenthinning treatment units can be detectedeven over this relatively short period of time(five growing seasons). Height growth wasless pronounced in the mature (age 75 years)

Figure 2. (A) Orthophotograph of selected area (courtesy of Washington Department of Natural Resources), and (B) individual tree-levelsegmentation of the LIDAR canopy height model via morphological watershed algorithm (color-coded by height; black lines indicateboundaries around crowns).

Figure 3. Comparison of 1999 and 2003 LIDAR crown measurements in a heavily thinnedstrip of mature forest in the Capitol State Forest study area. (From top to bottom) 1999orthophotograph; profile view of all 1999 LIDAR points (color-coded by heightaboveground) measured within the yellow box shown in the orthophotograph; plan viewof all 1999 points; and, plan view of all 2003 points. Note the crown expansion between1999 and 2003 that is apparent in the red square and the tree that was removed (becauseof windthrow) apparent in the red circle.


heavily thinned unit (approximately 0–2m), where the primary response to the treat-ment was increased crown expansion, thanin the mature unthinned control unit, wherethe height growth was in the range of 1–3 m(Figure 4). Not surprisingly, the heightgrowth within a younger (age 35 years)stand was much higher (approximately 3–5m) than in the mature stands. The capabilityof LIDAR to measure accurately the growthrates of individual dominant and codomi-nant trees across an entire forest clearly pro-vides an opportunity for much more accu-rate and spatially explicit assessment of sitequality and growth analysis.

Plot-Level LIDAR-Based ForestStructure Measurement

The basic principles of allometry, or lawsof proportional growth, can be used to quan-titatively model the relationship between thedimensions of various components of a forestsystem, including canopy height, biomass,basal area, and foliar surfaces (West et al.1997). These principles can be used to developregression models relating the spatial distribu-tion of LIDAR returns within a plot area toplot-level stand inventory variables (e.g.,height, volume, stocking, and basal area) be-cause LIDAR measurements essentially repre-sent a detailed measurement of all reflectingsurfaces within a canopy volume (foliage,branches, and stems). This approach is appro-priate when LIDAR data are collected at alower density (i.e., 1- to 2-m spacing betweenpoints) or the vertical structure of the forest iscomplex (i.e., composed of multiple canopystrata, perhaps with a significant understory

component). The metrics used to describe thespatial distribution of LIDAR returns in a plotarea include height percentiles, mean height,maximum height, coefficient of variation ofheight, and a LIDAR-derived measure of can-opy cover (e.g., percentage of LIDAR first re-turns above 2 m). This plot-level approach hasbeen used by researchers in North Americaand Europe to estimate stand inventory pa-rameters in several different forest types, wherepredictive regression models were shown to ex-plain from 80 to 99% of the variation (i.e., R2)in field-measured values (Means et al. 2000,Næsset and Økland 2002, Lim and Treitz2004). In a study performed using 99 fieldplots in second-growth Douglas-fir (P. men-ziesii (Mirb.) Franco var. menziesii) measuredat the Capitol State Forest study area in Wash-ington State, strong regression relationshipsbetween LIDAR-derived predictors and field-measured values were found for several criticalinventory parameters, including basal area (R2

� 0.91), stem volume (R2 � 0.92), dominantheight (R2 � 0.96), and biomass (R2 � 0.91;Andersen et al. 2005a). Because this approachrelies on a single mathematical model to relatethe LIDAR metrics to a given inventory pa-rameter over a range of different stand types, itis important to obtain representative plot-levelfield data that capture the full range of variabil-ity present in the area of LIDAR coverage. Re-cent research in the West Virginia mixed hard-wood forests also has indicated that theintensity data (sometimes referred to as “reflec-tance”) of the NIR reflection from LIDARdata acquired in leaf-off conditions are usefulfor some hardwood species classificationswhen used in conjunction with LIDAR geo-

metric data (Brandtberg et al. 2003). Theseintensity data likewise can be helpful to dis-tinguish between live and dead crowns whenLIDAR data are collected during leaf-onconditions.

Another promising application ofLIDAR technology to forest inventory is in thearea of canopy fuel mapping (Riano et al.2004). Resource managers rely on accurateand spatially explicit estimates of forest canopyfuel parameters, including canopy cover, can-opy height, crown bulk density, and canopybase height to support fire behavior modelingand fuel mitigation programs. In a study per-formed at the Capitol State Forest, regressionanalysis was used to develop strong predictivemodels relating a variety of LIDAR-based for-est structure metrics to plot-level canopy fuelestimates derived from field inventory data[sqrt(crown fuel weight), R2 � 0.86; ln(crownbulk density), R2 � 0.84; canopy baseheight, R2 � 0.77; canopy height, R2 � 0.98(Andersen et al. 2005b)]. These regressionmodels then can be used to generate digital

Figure 4. LIDAR-based measurement of individual conifer tree growth (1999–2003). (A) Selected area within Capitol State Forest study areashown in LIDAR canopy height model color-coded by canopy height, and (B) LIDAR-derived individual tree height growth measurementcolor-coded by height growth. A significant difference in height growth between stands is evident [control (75-year-old, unthinned stand)approximately 1–3 m in growth; young (35-year-old stand) approximately 3–5 m; heavily thinned (75-year-old stand) approximately 0–2m]. Segments colored white indicate hardwoods that were excluded from this conifer growth analysis.

Figure 5. LIDAR-derived canopy fuel weightmap (30-m resolution), Capitol State Foreststudy area (Andersen et al. 2005b).


maps of canopy fuel parameters over the ex-tent of the LIDAR coverage. Canopy fuelweight, e.g., can be mapped over the land-scape (Figure 5).

Need for LIDAR MissionStandards and Specifications

Today, we are in a position withLIDAR technology similar to where ourpredecessors were with aerial photographyin the early part of the last century. By1930, it was obvious that aerial photogra-phy was providing new data on the extent,composition, and volume of forests, aswell as information for many other naturalresource management activities; yet ittook many more years for agencies to de-velop flight specifications and cooperative,cost-sharing agreements to allow periodicwide-area photography missions. It is in-creasingly evident that LIDAR provides3D geo- and biospatial data at an unprec-edented level of detail and accuracy, butstandards and specifications have not beenestablished for collecting LIDAR datasuitable for use in a wide range of naturalresource management activities. At thesame time, many large LIDAR projects(county- and statewide acquisitions) arebeing flown by a multitude of local, state, andfederal agencies for single-use managementneeds (e.g., flood risk mapping, updated digi-tal elevation models (DEM), or geologic faultdetection), often without consideration as tohow the data might be used for forest vege-tation measurements and monitoring. ManyLIDAR data sets, for instance, are being flownwithout collecting (or without requiring deliv-ery to the client) return intensity informationthat is very useful for discerning forest types oridentifying mortality, and in some cases, spe-cies differences. Furthermore, many contractsalso have not required delivery of all returns—they simply specify bare-ground DEMs or afiltered subset of the data that only includesground points. Vegetation data often are lostor must be repurchased from the vendor.

There is an immediate need to startdeveloping standards and specificationsfor LIDAR data collections so that data aremore widely available for use by local,state, and federal natural resource man-agement agencies. Again, the multiagencyworking groups and agreements estab-lished to organize the collection and dis-tribution of periodic aerial photographyprovide models for how coordinatedLIDAR projects could be planned and fi-

nanced to insure that future data acquisi-tions meet multiresource needs. The FederalEmergency Management Agency (FEMA)has taken the lead in establishing guidelinesand specifications for LIDAR terrain map-ping for flood hazard mapping (FEMA2003); however, there has not been a similarcoordinated effort between natural resourcemanagement agencies. The FEMA stan-dards provide a good starting point onwhich to build more comprehensive stan-dards that meet multiresource needs.

Standards and specifications are neededfor LIDAR missions (sensor settings and flightspecifications) and for delivered products.There has been limited research on neededflight and sensor specifications. Evans et al.(2001) proposed research to examine the ef-fects of LIDAR flight and sensor specificationsover forests of differing density and spatial ar-rangements. A more coordinated, comprehen-sive research effort is needed to develop datacollection standards and specifications over amore complete range of forest conditions. Inaddition, standards are needed for products ofLIDAR missions to insure their usefulness forforest measurements.

To stimulate development of bothtypes of standards and specifications, thereare several simple, easily understood andwidely recognized LIDAR-derived forestmapping products that many agencies andspecialists within organizations would finduseful. The following five could be gener-

ated easily, assuming consistent data collec-tion standards are implemented:

1. High-resolution (1–5 m) bare-groundDEM. These DEMs provide improveddata for many applications including hy-drologic and erosion process modeling,landscape modeling, road and harvestplanning and design, and geographic in-formation system analysis (Figure 6B).

2. Canopy height models (CHM). CHMsprovide spatially explicit stand structuredata over the landscape for estimation ofgrowing stock, input for habitat and firemodels, and any other resource planningactivities where spatial arrangement andtree height are important considerations(Figure 6C).

3. Canopy cover maps. These images pro-vide a direct measurement of cover byheight aboveground. Figure 6D illus-trates canopy cover where canopy heightis greater than 2 m.

4. LIDAR intensity images. These high-resolution images can be matched withexisting orthophotographs and otherdigital imagery for change detection andmonitoring over time. They also are use-ful in verifying the registration of LIDARdata with other geospatial data layers. Asshown in Figure 6D, intensity data canbe used in conjunction with CHMs toidentify hardwood (brown) and conifercanopy areas (green).

5. All returns data set. This archive of all the

Figure 6. Comparison of a traditional color orthophotograph to LIDAR-derived images forthe same area: (A) orthophotograph; (B) bare-ground DEM; (C) CHM (canopy height is lessthan 2 m in gray areas); and (D) canopy cover image colored by leaf-offLIDAR intensity, where brown low-intensity areas indicate hardwood cover and greenhigh-intensity areas indicate conifer cover.


LIDAR returns and their associated re-flectance intensity could be used for awide range of specialized analysis andprovides baseline data on current terrainand vegetation structure that could beused in the future for change detectionand monitoring (e.g., crown expansionor dieback). At a minimum, this data setshould include pulse number, returnnumber, east coordinate, north coordi-nate, elevation, and return intensity foreach LIDAR return and metadata docu-menting the LIDAR mission flight pa-rameters, sensor type and settings, GPScontrol, horizontal and vertical datum,coordinate units and projection, and dateand time of mission. Ideally, all returndata files should be in the American So-ciety for Photogrammetry and RemoteSensing LIDAR data exchange format.

Leveraging Ongoing GroundPlot Measurements

Several projects have reported excellentresults using LIDAR in double-samplingforest inventory approaches (Næsset 2002,Parker and Evans 2004). The Forest Inven-tory and Analysis program (FIA), USDAForest Service, is continually measuring per-manent sample plots in the United Statesthat could be used over time to develop ro-bust LIDAR regression estimators for majorcanopy variables. However, for FIA plots tobe useful in double-sampling regressionanalysis, plot locations will need to be mea-sured more accurately than is the currentpractice (nominally 10 m, but likely 10–50m). Ideally, locations should be accurate towithin a meter (well within the capabilitiesof differentially corrected GPS surveys) toallow LIDAR point clouds to be aligned cor-rectly with ground plots. Ironically, with thedevelopment of highly precise direct georef-erencing systems for airborne sensors, now,it is often more difficult to obtain accurateGPS ground positions, because of canopyinterference with GPS reception, than it is togeoreference airborne remote sensing data.Given the large ongoing investment that isbeing made in remeasurement of groundplots, it would seem that a small proportionof these ground plots (carefully selected tocover a wide range of forest stand condi-tions) should be located more carefully.LIDAR data could then be collected duringthe same season over these plots. Within afew years, an extensive archive of spatiallyaligned ground and LIDAR plot data would

be available for development of regressionmodels. These regressions would then beavailable for use with any large-area LIDARdata set (past or future) to estimate forestinventory parameters or other vegetationvariables for use in a multitude of land-man-agement exercises. This would provide avaluable method for spatially explicit moni-toring of forest change with unprecedentedaccuracy and resolution.

ConclusionsOver the last 5 years, numerous studies

have shown that LIDAR data can providehigh-resolution biospatial data for multire-source management and analyses includingtraditional forest inventory and more spe-cialized single-use analysis (e.g., canopy fuelestimates for fire behavior modeling). Si-multaneously, LIDAR has emerged as theleading technology for high-resolution ter-rain mapping, spurring the development ofnational guidelines and standards in this do-main. It appears there is a similar need todevelop national standards and guidelinesfor LIDAR data collection for forest vegeta-tion measurement and monitoring to insurethat the maximum value can be returnedfrom future LIDAR projects over forestedregions. Focusing attention on such stan-dards may also encourage LIDAR manufac-turers to modify scanners for more optimalsensing of vegetation, rather than simply ter-rain, particularly given the need for moni-toring forest change. Finally, the usefulnessand value of ongoing permanent groundsample plot measurements could be lever-aged by collecting more accurate plot loca-tions and collecting limited sets of LIDARdata over these plots.

Literature CitedANDERSEN, H.-E. 2003. Estimation of critical for-

est structure metrics through the spatial analysis ofairborne laser scanner data. Ph.D. dissertation,Univ. of Washington, Seattle, WA. 176 p.

ANDERSEN, H.-E., R.J. MCGAUGHEY, AND S.E.REUTEBUCH. 2005A. Forest measurement andmonitoring using high-resolution airborne li-dar. P. 109–120 in Productivity of western for-ests: A forest products focus, Harrington, C.A.,and S.H. Schoenholtz (eds.). UDSA For. Sev.Gen. Tech. Rep. PNW-GTR-642, PacificNorthwest Res. Sta., Portland, OR.

ANDERSEN, H.-E., R.J. MCGAUGHEY, AND S.E.REUTEBUCH. 2005B. Estimating forest canopyfuel parameters using LIDAR data. RemoteSens. Environ. 94(4):441–449.

BRANDTBERG, T., T.A. WARNER, R. LANDEN-BERGER, AND J. MCGRAW. 2003. Detectionand analysis of individual leaf-off tree crownsin small footprint, high sampling density

LIDAR data from the eastern deciduous forestin North America. Remote Sens. Environ.85(3):290–303.

EVANS, D.L., S.D. ROBERTS, J.W. MCCOMBS,AND R.L. HARRINGTON. 2001. Detection ofregularly spaced targets in small-footprint lidardata—research issues for consideration. Photo-gramm. Eng. Remote Sens. 67(10):1133–1136.

FEDERAL EMERGENCY MANAGEMENT AGENCY.2003. Appendix A: Guidance for aerial map-ping and surveying P. A39–A57 in Guidelinesand specifications for flood hazard mapping part-ners. Available online at www.fema.gov/pdf/fhm/frm_gsaa.pdf; last accessed Sept. 2005.

KRAUS, K., AND N. PFEIFER. 1998. Determinationof terrain models in wooded areas with air-borne laser scanner data. ISPRS J. Photo-gramm. Remote Sens. 53:193–203.

LIM, K.S., AND P.M. TREITZ. 2004. Estimation ofabove ground forest biomass from airbornediscrete return laser scanner data using cano-py-based quantile estimators. Scand. J. For.Res. 19(6):558–570.

MCCOMBS, J.W., S.D. ROBERTS, AND D.L.EVANS. 2003. Influence of fusing lidar andmultispectral imagery on remotely sensed esti-mates of stand density and mean tree height ina managed loblolly pine plantation. For. Sci.49(3):457–466.

MEANS, J.E., A.A. ACKER, B.J. FITT, M. REN-SLOW, L. EMERSON, AND C. HENDRIX. 2000.Predicting forest stand characteristics with air-borne scanning LIDAR. Photogramm. Eng. Re-mote Sens. 66(11):1367–1371.

NÆSSET, E. 2002. Predicting forest stand charac-teristics with airborne laser using a practicaltwo-stage procedure and field data. RemoteSens. Environ. 80(1):88–99.

NÆSSET, E., AND T. ØKLAND. 2002. Estimatingtree height and tree crown properties using air-borne scanning laser in a boreal nature reserve.Remote Sens. Environ. 79(1):105–115.

NÆSSET, E., T. GOBAKKEN, J. HOLMGREN, H.HYYPPA, J. HYYPPA, M. MALTAMO, M. NILS-SON, H. OLSSON, Å. PERSSON, AND U. SODER-MAN. 2004. Laser scanning of forest resources:The Nordic experience. Scand. J. For. Res.19(6):482–499.

OLSSON, H., AND E. NÆSSET (EDS.). 2004. Specialissue—ScandLaser workshop in Umeå, Swe-den, Sept. 2–4, 2003. Scand. J. For. Res. 19(6):481–582.

PARKER, R.C., AND D.L. EVANS. 2004. An appli-cation of LiDAR in a double-sample forest in-ventory. West. J. Appl. For. 19(2):95–101.

PEREIRA, L., AND L. JANSSEN. 1999. Suitability oflaser data for DTM generation: A case study inthe context of road planning and design. IS-PRS J. Photogramm. Remote Sens. 54:244–253.

PERSSON, Å., J. HOLMGREN, AND U. SODERMAN.2002. Detecting and measuring individualtrees using an airborne laser scanner. Photo-gramm. Eng. Remote Sens. 68(9):925–932.

POPESCU, S., R. WYNNE, AND R. NELSON. 2003.Measuring individual tree crown diameterwith lidar and assessing its influence on forestvolume and biomass. Can. J. Remote Sens.29(5):564–577.


POPESCU, S.C., R.H. WYNNE, AND J.A. SCRIVANI.2004. Fusion of small-footprint lidar and mul-tispectral data to estimate plot-level volumeand biomass in deciduous and pine forests inVirginia, USA. For. Sci. 50(4):551–565.

REUTEBUCH, S.E., R.J. MCGAUGHEY, H.-E.ANDERSEN, AND W.W. CARSON. 2003. Accu-racy of a high-resolution lidar terrain modelunder a conifer forest canopy. Can. J. RemoteSens. 29(5):527–535.

RIANO, D., E. CHUVIECO, S. CONDIS, J. GONZA-LEZ-MATESANZ, AND S.L. USTIN. 2004. Gener-ation of crown bulk density for Pinus sylvestrisfrom LIDAR. Remote Sens. Environ. 92(3):345–352.

SCHARDT, M., M. ZIEGLER, A. WIMMER, R.WACK, AND J. HYYPPA. 2002. Assessment offorest parameters by means of laser scanning.Int. Arch. Photogramm. Remote Sens. XXXIV-(part 3A):302–309.

SCHREUDER, H., T. GREGOIRE, AND G. WOOD.1993. Sampling methods for multiresource forestinventory. Wiley, New York. 464 p.

STRAUB, B.-M. 2003. A top-down operator forthe automatic extraction of trees—concept

and performance evaluation. Int. Arch. Photo-gramm. Remote Sens., XXXIV (part 3/W13):34–39.

WEHR, A., AND U. LOHR. 1999. Airborne laserscanning-an introduction and overview. IS-PRS J. Photogramm. Remote Sens. 54:68–82.

WEST, G.B., J.H. BROWN, AND B.J. ENQUIST.1997. A general model for the origin of allo-metric scaling law in biology. Science 276:122–126.

WULDER, M.A., P. NORMAN, AND C. WITTE

(EDS.). 2003. Special issue: Lidar remote sens-ing of forest structure and terrain. Can. J. Re-mote Sens. 29(5):527–678.

YU, X., J. HYYPPA, H. KAARTINEN, AND M. MAL-TAMO. 2004. Automatic detection of harvestedtrees and determination of forest growth usingairborne laser scanning. Remote Sens. Environ.90:451–462.

ZIEGLER, M., H. KONRAD, J. HOFRICHTER, A.WIMMER, G. RUPPERT, M. SCHARDT, AND J.HYYPPA. 2000. Assessment of forest attributesand single-tree segmentation by means of laserscanning. In Laser Radar Technology and Appli-cations V, Kamerman, G.W., U.N. Singh, C.

Werner, and V.V. Molebny (eds.). Proc. ofSPIE 4035:73–84.

Stephen E. Reutebuch ([email protected])is team leader, USDA Forest Service, PacificNorthwest Research Station, 292 Bloedel Hall,University of Washington, Seattle, WA98195-2100. Hans-Erik Andersen ([email protected]) is research scientist, Univer-sity of Washington, Precision Forestry Coopera-tive, 292 Bloedel Hall, University ofWashington, Seattle, WA 98195-2100. RobertJ. McGaughey ([email protected]) is re-search forester, USDA Forest Service, PO Box352100, University of Washington, Seattle, WA98195-2100. The authors acknowledge theWashington Department of Natural Resources,the Joint Fire Science Program, and the Univer-sity of Washington, Precision Forestry AdvancedTechnology Initiative for their generous support ofour LIDAR measurement research efforts.


Light Detection and Ranging (LIDAR): An Emerging Tool for ...

Documents

Transcript of Light Detection and Ranging (LIDAR): An Emerging Tool for ...