Asd

Ja

b

ARRA

KPTLGEE

U

l

0h

Agricultural and Forest Meteorology 180 (2013) 86– 96

Contents lists available at SciVerse ScienceDirect

Agricultural and Forest Meteorology

jou rn al hom epage : www.elsev ier .com/ locate /agr formet

lightweight, low cost autonomously operating terrestrial lasercanner for quantifying and monitoring ecosystem structuralynamics

an U.H. Eitel a,b,∗, Lee A. Vierlinga,b, Troy S. Magneya

Geospatial Laboratory for Environmental Dynamics, University of Idaho, Moscow, ID 83844-1135, USAMcCall Outdoor Science School, University of Idaho, McCall, ID 83638, USA

a r t i c l e i n f o

rticle history:eceived 21 June 2012eceived in revised form 24 May 2013ccepted 26 May 2013

eywords:lant structureLSiDARround validationnvironmental changecosystem dynamics

a b s t r a c t

The three-dimensional (3-D) structure of ecosystems is inherently dynamic. However, this is oftenignored in ecological studies because it is difficult to characterize using traditional field methods. Terres-trial laser scanning (TLS) is a rapidly maturing technique to complement and enhance traditional fieldmethods for quantifying 3-D geometric properties of ecosystems. Two major limitations of TLS includethe low temporal resolution that often exists between each data acquisition, and the relatively high cost ofsuch systems (entry level systems cost >$40,000 USD) that puts this method out of reach for many poten-tial users. Consequently, TLS is currently limited as a mainstream method for capturing 3-D geometricecosystem dynamics. The objectives of this study were to (i) describe the design of a lightweight (3.85 kg),low-cost ($<12,000 USD), autonomously operating terrestrial laser scanner (ATLS) and to (ii) test its abil-ity to provide data to quantify and monitor ecological characteristics that exhibit structural change. Wetested the utility of the ATLS data to quantify plant growth by measuring plants with different heights anddiameter at breast height (DBH). Specifically, we derived the canopy heights of a conifer tree (Engelmannspruce, Picea engelmannii), broadleaf tree (Quaking aspen, Populus tremuloides), graminoid (Calamagrostisx acutiflora), and forb (Hemerocallis lilioasphodelus), and the DBH of Ponderosa Pine (Pinus ponderosa) andDouglas-fir (Pseudotsuga menziesii) trees. The ATLS was also tested under varying weather conditions(including rain, snowfall and temperature ranging from −9.1 to 21.1 ◦C), to quantify canopy structuralchanges in quaking aspen during leaf drop relative to a Ponderosa Pine that retained its leaves over thesame time period. We also compared canopy structural changes quantified by ATLS canopy laser returnswith those quantified using a commercial TLS. Our results showed strong agreements between observedand ATLS derived conifer tree canopy height (RMSE = 0.96 cm, r2 = 1.00, slope = 0.96, intercept = 1.43),broadleaf tree canopy height (RMSE = 0.08 m, r2 = 0.99, slope = 1.01, intercept = −0.38), graminoid andforb canopy height (RMSE = 1.56 cm, r2 = 0.98, slope = 1.04, intercept = −2.22), and DBH (RMSE = 2.24 cm,r2 = 0.99, slope = 0.99, intercept = 0.45). A strong relationship (r2 = 0.86) also existed between the num-

ber of TLS and ATLS canopy laser returns. Our results indicate that the ATLS is suitable for monitoringand quantifying dynamics of plant growth and potentially many other 3-D properties of ecosystems.While further research is needed to better understand the effect of scan resolution, beam divergence,and atmospheric conditions on the accuracy of ATLS derived metrics, this instrument has great promisefor providing new insights into dynamic ecosystem processes that are currently difficult to monitor athigh temporal and spatial resolution.

∗ Corresponding author at: Geospatial Laboratory for Environmental Dynamics,niversity of Idaho, Moscow, ID 83844-1135, USA.

E-mail addresses: jeitel@uidaho.edu, jeitel@vandals.uidaho.edu (J.U.H. Eitel),eev@uidaho.edu (L.A. Vierling), tmagney@uidaho.edu (T.S. Magney).

168-1923/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.agrformet.2013.05.012

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

Ecosystem three-dimensional (3-D) structure is inherentlydynamic, often changing in complex, non-linear ways as a resultof both sudden (i.e., “pulsed”) and continual (“pressed”) envi-

ronmental influences (Ives and Carpenter, 2007). Methodologicaltechniques that allow for ecosystem and geomorphological struc-tural changes to be quantified and monitored at high temporaland spatial resolutions are therefore needed to better understand

d Fore

dflteofSVlEis

sesprshovR3tcSsposaAde2

tqlt22(WsGegtbtEsurapitrssspb

J.U.H. Eitel et al. / Agricultural an

ynamic structural processes such as plant growth and decay,uvial, lacustrine, and coastal geomorphology, snow/ice accumula-ion and melt, and other aspects of landscape evolution (e.g., Starekt al., 2011). New understanding gained from such studies could bef great practical and scientific significance, ranging in applicationrom tracing plant growth rate response to climate change (e.g.,churr et al., 2006), to tracking 3-D changes in wildlife habitat (e.g.,ierling et al., 2008), to untangling complexities of landform evo-

ution (Lim et al., 2005; Roncat et al., 2011), among many others.mpirical data collected through these efforts would provide newnformation upon which to develop and validate mechanistic 3-Dimulation models that are dynamic in time.

Although 3-D geometric dynamics of ecosystems have beentudied and modeled for decades (e.g., Shugart et al., 1973; Ryut al., 2012), relatively little quantitative 3-D data on ecosystemtructural change exists. Airborne and satellite RADAR and LiDARlatforms have bolstered understanding of many critical topicselating to ecosystem and geomorphological structure at the land-cape scale (e.g., Treuhaft and Cloude, 1999; Lefsky et al., 2002);owever, satellite systems operate at relatively coarse spatial res-lution and airborne LiDAR datasets are generally acquired at aery low temporal frequency. Therefore, both airborne and satelliteADAR and LiDAR may not be suitable to study some topics where-D structural change occurs rapidly, or at a fine spatial scale. Athese fine spatiotemporal scales, quantification of 3-D geometrichanges traditionally relies on manual field measurements (e.g.,churr et al., 2006). Unfortunately, by their nature, manual mea-urements often suffer from several disadvantages: they may haveoor spatial resolution and/or limited spatial extent, they may maskr influence the process of interest by altering the original studyite (e.g., by removing biomass or simply accessing the study site),nd they are often laborious and costly (e.g., Jester and Klik, 2005).s a consequence, our current understanding of 3-D ecosystemynamics is limited and hinders our ability to model and predictcosystem responses to changing environmental conditions (Arora,002; Scanlon et al., 2005; Schurr et al., 2006).

Terrestrial laser scanning (TLS) is a rapidly maturing techniquehat may complement and enhance traditional field methods foruantifying structural properties of ecosystems at the fine (cm-

evel) scale. For example, terrestrial laser scanners have been usedo quantify the 3-D geometric properties of plants (Clawges et al.,007; Rosell et al., 2009; Eitel et al., 2010; Keightley and Bawden,010; Moorthy et al., 2011; Vierling et al., 2012), soil surfacesHaubrock et al., 2009; Eitel et al., 2011b; Sankey et al., 2011;

enske et al., 2012; Hancock et al., 2008; Perroy et al., 2010), snowurfaces (Egli et al., 2012; Prokop, 2008; Schaffhauser et al., 2008;utmann et al., 2011), stream banks and stream channels (Milant al., 2007; Williams et al., 2011), cliffs (Lim et al., 2005), andlaciers (Schwalbe et al., 2008). However, one of the major limi-ations of TLS is the low temporal resolution that generally existsetween repeated data acquisitions which limits its use to quan-ify 3-D geometric changes (Milan et al., 2007; Staley et al., 2011;gli et al., 2012). For example, Milan et al. (2007) showed that ero-ion and deposition volumes in a proglacial river were increasinglynderestimated with a progressively coarser temporal samplingesolution. An 8-day surveying interval revealed 67% less erosionnd 14% less deposition than a daily survey interval. The low tem-oral resolution that often exists between TLS data acquisitions

s mainly due to the fact that it is expensive and labor intensiveo conduct repeated surveys. Repeated measurements may alsoequire the set-up of spatially invariable targets to tie multiple TLSurveys to a common datum, which can be time consuming and

ometimes difficult, dangerous, and impractical. In addition, TLSystems are often quite expensive to purchase (e.g., an entry-levelrice of ∼$40,000 USD), therefore limiting their accessibility to aroad user base. Finally, many TLS systems are often too heavy to

st Meteorology 180 (2013) 86– 96 87

be mounted on infrastructure such as meteorological or flux tow-ers. Because autonomous collection of ecosystem structure datawould well complement such meteorological and/or flux measure-ments, lightweight instruments (e.g., <4 kg) would enable a suiteof new applications for integration with common meteorologicalinstrumentation.

An alternative to traditional TLS instruments is to assemble aterrestrial laser system from off-the shelf items to autonomouslyand continuously scan a study site. Gutmann (2010) and Gutmannet al. (2011), for example, have pioneered work in this area, using alaser rangefinder and pan-tilt unit to monitor 3-D snow accumula-tion and melt dynamics. To date, however, a thorough descriptionof such a system is missing in the scientific literature.

Here, we provide detailed explanation of such a system – here-after referred to as an autonomously operating terrestrial laserscanner (ATLS). Also, because relatively little is known about theability of such an autonomous system to characterize structuralproperties of surfaces other than snow (e.g., quantifying plantdynamics), we present additional analyses in this regard. Comparedto monitoring structural properties of continuous surfaces such assnow, quantifying discontinuous surfaces such as plant canopiescan be complicated when using a laser rangefinder because edgesof leaves and branches can split a single laser pulse so it may striketwo or more objects (e.g., multiple leaf surfaces, or leaf and branchsurfaces) (Hebert and Krotkov, 1992; Tuley et al., 2004). If the splitlaser beam returns to the laser rangefinder from the front and back-ground object, the instrument calculates a single distance value byintegrating the distances to the front and background object pro-portional to their signal strength (Hebert and Krotkov, 1992; Tuleyet al., 2004; Rosell et al., 2009; Eitel et al., 2010; Sanz-Cortiella et al.,2011a). This then results in a mixed edge return (also known asmixed pixel, ghost return, or air return) inherent to all laser basedranging methods, where the recorded distance is neither the dis-tance to the front and background object but rather the distanceto a phantom object (i.e., fictitious point) that lays somewhere inbetween both objects (Hebert and Krotkov, 1992; Tuley et al., 2004).Also, the reflective properties of vegetation differ from snow whichmay affect the accuracy of laser derived metrics and should beconsidered when assembling an ATLS.

Our objective in this study was therefore first to describe thedesign of an autonomously and continuously operating terrestriallaser system (ATLS), and second to conduct testing on our ability touse an ATLS to quantify 3-D dynamics of plant canopies.

2. Methods

2.1. Autonomously operating terrestrial laser scanner (ATLS)instrument design

The laser system components consist of a time-of-flight laserrangefinder, circular level, electronic pan-tilt unit, tribrach, data-logger, and power supply (battery and 70 W solar panel) (Fig. 1).At air temperatures above freezing, the ATLS requires 12 V DCpower and 0.5 amp of current. It is important to note that thepower supply was designed to ensure a continuous operation ofthe ATLS for sun hours and temperatures typical for Idaho dur-ing the growing season. Hence, the power supply might need tobe adjusted (e.g., increase number of solar panels and/or theirnominal maximum power) for areas with different temperaturesand/or sun hours to guarantee an uninterrupted power supply tothe ATLS. The cost for all ATLS components as of 2013 is $11,841

USD (Table 1). Depending on the objectives of the study, proprietarysoftware might be necessary which might add to the overall cost.However, open-source software packages are available such as LAS-tools (Isenburg, 2007–2012), GEON Points2Grid Utility (Kim et al.,

88 J.U.H. Eitel et al. / Agricultural and Fore

Fpo

2tPDftt

pwaa

The ILR 1191 employs a near infrared (905 nm) laser and has

TA

T

ig. 1. Autonomously operating laser scanner (ATLS) and its main off-the-shelf com-onents, including the laser rangefinder, pan-tilt unit (PTU), and tribrach mountedn a tripod.

006), LViz (Conner, 2010) or R (R Development Core Team, 2011)hat might be sufficient to post-process the ATLS data. The GEONoints2Grid Utility allows the user to convert laser point clouds toigital Terrain Models (DTM). The LViz tool as well as the plot3-D

unction in the R package “rgl” (Adler and Murdoch, 2011) allowshe user to visualize the laser point cloud in 3-D. LAStools allowso visualize, clip, and process TLS data.

Laser rangefinders suitable for an ATLS should ideally be (i) inex-

ensive, (ii) able to range distances >20 m, and (iii) operate under aide range of weather conditions (e.g., dry and wet conditions, high

nd low air temperatures). The most widely commercially avail-ble laser rangefinders that fulfill these requirements are phase

able 1pproximate costs for ATLS as of 2013 (in U.S. dollars).

Item Provider

ILR1191 laser distancesensor

Micro-Epsilon Messtechnik GmbH& Co. KG, 94496 Ortenburg,Germany

PTU-46–17 W pan tiltunit

FLIR Motion Control Systems, Inc.,Burlingame, CA 94010, USA

CST/Berger “ValueLine” Tribrach

Tiger Supplies Inc., 27 Selvage St.,Irvington NJ 07111, USA

Seco Fixed TribrachAdapter

Tiger Supplies Inc., 27 Selvage St.,Irvington NJ 07111, USA

CR1000 datalogger Campbell Scientific, 815 West 1800North, Logan, UT 84321-1784, USANL115 Ethernet

Interface &CompactFlashModule fordatalogger

2G CompactFlashMemory Card

PC400 dataloggersupport software(includes CRBasicprogramminglanguage)

Power supply (battery,70 W solar panel,Morning StarSunSaver-10 10 A12 V Regulator)

Custom laser mount University of Idaho machine shop,Moscow, ID 83844, USA

otal: $11,841 USD.

st Meteorology 180 (2013) 86– 96

based laser rangefinders and discrete return, time-of-flight laserrangefinders. To calculate a distance to the target of interest, phasebased laser rangefinders use the phase shift between the sinu-soidally modulated laser light that is transmitted and received bythe sensor. Discrete return, time-of-flight laser rangefinders calcu-late a distance based on the time-of-flight (t) of a laser pulse that isreflected back to the senor from a surveyed object (distance = (ct)/2,where c is the speed of light and t is round-trip elapsed time of lightpropagation).

We chose a Class 1 (harmless to the eyes), discrete return,time-of-flight laser rangefinder (optoNCDT ILR 1191, Micro-EpsilonMesstechnik GmbH & Co. KG, Ortenburg, Germany) designed tooperate in harsh outdoor environments (e.g., has internal heater,operation temperature ranges between −40 ◦C and +60 ◦C). TheILR 1191 can measure up to 2000 distance readings per second(2000 Hz). The rangefinder was programmed to record each singledistance reading as opposed to recording an average of two or moredistance readings which is undesirable when using rangefinders forcharacterizing plant canopies (Parker et al., 2004). Further, the ILR1191 was programmed to record a value of zero in case of a ‘skyhit’ where no object is in the path of the laser beam or the intensityof the returned laser pulse is below a laser specific threshold. Thebeam divergence of the IRL 1191 is 1.7 mrad, resulting in a beamdiameter of 45 mm at 10 m, 155 mm at 75 m, 240 mm at 125 m,and 538 mm at 300 m (Instructional Manual optoNCDT ILR 1191,Micro-Epsilon Messtechnik GmbH & Co. KG, Ortenburg, Germany).The light spot geometry (height and width) of the laser beam isrectangular and changes with distance as follows: 45 mm × 40 mmat 10 m, 155 mm × 105 mm at 75 m, 240 mm × 155 mm at 125 m,and 538 mm × 360 mm at 300 m (personal communication, JoshJones, Micro-Epsilon). The power distribution within the laser beamcross-section is uniform.

a maximum range of 150 m, 300 m, and 500 m for surfaces with areflectance at 905 nm of 6%, 10%, and 90%, respectively. As opposedto lasers that operate in the visible wavelengths, a NIR laser ensures

Date quoted Qty Amount (USD)

1/20/2012 1 $5575

1/20/2012 1 $2765

4/20/2012 1 $100

4/20/2012 1 $33

1/19/2012 1 $13831 $274

1 $87

1 $303

1 $1221

5/24/2012 1 $100

d Fore

af(vIlwG

eBooTvvmcc5agpSg

(

(

(((

(((

sssst

T

wptt

dsc

x

y

z

tl

J.U.H. Eitel et al. / Agricultural an

strong laser return signal from both vegetated and bare soil sur-aces. Vegetation strongly absorbs red light in contrast to NIR lightGausman, 1985). Soils also reflect stronger in the NIR than in theisible part of the spectrum (e.g., Clark, 1999; Eitel et al., 2009).f a visible laser would be used for monitoring vegetation, theaser signal would be strongly absorbed by the plant’s chlorophyll

hich could adversely affect the point accuracy of leaf surfaces (Vannechten et al., 2008).

The laser rangefinder was mounted on an electronic, weath-rized, pan-tilt unit (PTU-D46-17W, FLIR Motion Systems,urlingame, CA, USA) that allows datalogger controlled positioningf the laser rangefinder to a given pan (azimuth) and tilt (zenith)rientation. The combined mass of the laser/pan-tilt unit is 3.85 kg.he pan-tilt unit (PTU) can travel to any orientation within 360◦

iew azimuth and 10–121◦ view zenith angle (i.e., where nadiriewing is 0◦ view zenith angle), in angular resolution step incre-ents as small as 0.013◦. A complete overview of pan-tilt unit

ommands for the PTU used in this study are given the pan-tilt unitommand reference manual (http://www.flir.com/mcs/view/?id=3707&collectionid=581&col=53708). To run the laser rangefindernd pan-tilt unit autonomously and to collect data, a datalog-er (CR1000, Campbell Scientific, Logan, UT, USA) was used androgrammed using the CRBasic programming language (Campbellcientific, Logan, UT, USA). The sequential flow of the CRBasic pro-ram is as follows;

1) Send command from datalogger to PTU to achieve a given panand tilt position

2) Send command from datalogger to PTU to query current panand tilt position information

3) PTU sends current pan-tilt position to datalogger4) Datalogger stores current pan-tilt position5) Send command from datalogger to laser rangefinder to collect a

single distance measurement (Note: mixed edge returns or ‘skyhits’ do not affect the stability of the CRBasic program or causemeasurement delays)

6) Laser rangefinder sends distance data to the datalogger7) Datalogger stores the distance information8) Repeat at new angle until area of interest is scanned

This can be easily adjusted, so a given area can be repeatedlycanned at a chosen time interval (e.g., daily or weekly). The CRBa-ic program is freely available upon request from the authors. Thecan duration for each survey point is approximately 1.5 s. The totalcanning time (Ts) for a survey area can be approximated based onhe following formula:

s =[

Panmax − Panmin

Hres× Tiltmax − Tiltmin

Vres

]× 1.5 s (1)

here Panmax is the maximum pan angle, Panmin is the minimuman angle, Tiltmax is the maximum tilt angle, Tiltmin is the minimumilt angle, Hres is the horizontal scan angular resolution, and Vres ishe vertical scan angular resolution.

For each survey point, the datalogger records the spherical coor-inates: distance (r), azimuth angle (ϕ), and zenith angle (�). Thepherical coordinates can then be used to calculate the Cartesianoordinates x, y, and z as follows:

= r cos(ϕ) sin(�) (2)

= r sin(ϕ) sin(�) (3)

= r cos(�) (4)

To ensure accurate x, y, and z coordinates, it is important to levelhe sensor before measurements start. For this, the PTU with theaser rangefinder has to be brought to a tilt position of 0◦ before

st Meteorology 180 (2013) 86– 96 89

adjusting the foot screws of the tribrach until the circular bubble islevel.

By changing the increment between consecutive pan and tiltangles within the CRBasic program, the horizontal and verticalpoint spacing can be changed. For example, by changing the incre-ment between consecutive pan angles from 2◦ to 1◦ and theincrement between consecutive tilt angles from 2◦ to 3◦, one wouldincrease the horizontal point spacing while decreasing the verticalpoint spacing. The angular measurement pattern of the laser scan-ner causes the point density to be inversely related to the distancebetween the scanner location and the target objects (Van der Zandeet al., 2006). For example, if one does have a horizontal scan reso-lution of 1◦, one would have a point spacing of 0.87 m at a distance50 m away from the ATLS, and a point spacing of 1.75 m at a dis-tance 100 m away from the ATLS. The vertical and horizontal pointspacing at a given distance can be calculated using basic geometry:

PSr = tan(0.5 × Sres) × 2 × r (5)

where PSr is the point spacing at a distance r and Sres is the verticalor horizontal scan resolution.

2.2. ATLS system validation

A set of experiments was conducted to test the suitability of theATLS to quantify and monitor 3D ecosystem change. The first set ofexperiments, hereafter referred to as basic testing, was conductedunder controlled laboratory conditions (e.g., no wind, constanttemperature, no rain) to test the ability of the ATLS to record useful3-D plant structural variability. The second experiment, hereafterreferred to as field testing, was conducted to test the suitability ofthe ATLS to monitor 3-D ecosystem dynamics under field conditions(e.g., rain, snow, wind, and high and low temperatures).

2.2.1. Basic testingWe harvested a total of five Engelmann spruce (Picea engel-

mannii) saplings. The tree heights were manually measured with ameter tape. The heights ranged between 28.58 and 88.9 cm, mim-icking sapling height growth. The trees were arranged next to eachother and mounted on a wooden board so that the canopies did notoverlap. The distance between the ATLS and the trees was approx-imately 3 m. The trees were scanned with a vertical and horizontalscan resolution of 11 mrad. To examine the mixed edge effect on theaccuracy of tree canopy height estimates, the trees were scanned asecond time with a solid object positioned 1 m behind the trees.

To test the ability of the ATLS to quantify and monitor the 3Dgrowth of graminoids and herbs, the canopy height of Feather ReedGrass (Calamagrostis x acutiflora) and Daylily (Hemerocallis lilioas-phodelus) was manually measured with a distance tape and scannedseven times. To mimic plant growth, the Feather Reed Grass andDaylily were shortened between each set of measurements. Thedistance between the laser and the Feather Reed Grass and Daylilywas approximately 3 m and the vertical and horizontal scan reso-lution was 11 mrad.

2.2.2. Field testingDiameter at breast height (DBH) is defined as the tree girth at

1.37 m above the ground. Manual DBH measurements of 10 treeswere taken with a metric fabric diameter tape at a flat locationwithin Ponderosa State Park (116.0906◦ E, 44.9271◦ N), located inWestern central Idaho, USA. Only the DBH of tree stems that couldbe completely seen from the scan center were manually measured

and used for the analysis. Manually measured DBH ranged from24.1 to 100.7 cm. Trees found within the scan area were PonderosaPine (Pinus ponderosa) and Douglas-fir (Pseudotsuga menziesii). TheATLS was set up at a height of 1.37 m above the ground, and a 360◦

90 J.U.H. Eitel et al. / Agricultural and Forest Meteorology 180 (2013) 86– 96

in asp

ss

dssF−s(ctS3damItsum1Thmklc

2

p

Fig. 2. Time-lapse camera images of study site documenting change

can was taken with a scan resolution of 5.5 mrad to obtain a cross-ectional scan map of the area.

To monitor and map changes in canopy structural propertiesuring leaf drop-off in quaking aspen (Populus tremuloides), ATLScans were acquired between October 12 and 29, 2012 (Fig. 2). Thetudy site was located within the University of Idaho’s Herald Nokesamily Experimental Forest, near McCall, ID, USA (44.894878◦;116.065968◦). In addition to quaking aspen, species found at the

tudy site included Ponderosa Pine (Pinus ponderosa), Douglas-firP. menziesii), and grand fir (Abies grandis). The mean annual pre-ipitation at the study site is 58–69 cm, with a mean annual airemperature of 2.8–5 ◦C and a frost free period of 65–80 days (Soilurvey Staff). The scan resolution (spacing) of the ATLS was set to

mrad resulting in a laser point spacing of 3 cm at 10 m. To vali-ate ATLS measurements, TLS measurements were also conductedt the field site on October 12, 18, 25, and 29, 2012. TLS measure-ents were taken with a Leica ScanStation2 (Leica Geosystems

nc., Heerbrugg, Switzerland) with the similar scan resolution ashe ATLS. The time-of-flight TLS employed has a green (532 nm)canning laser with a beam divergence of 0.15 mrad, a scan rate ofp to 50 kHz, a maximum sample density of <1 mm, and a maxi-um range of 134 m for surfaces with a reflectance at 532 nm of

8%. Distance accuracy is ±4 mm and position accuracy is ±6 mm.he TLS was placed directly next to the ATLS at the same viewingeight to ensure similar viewing geometry between both instru-ents (Figs. 2 and 3). The tripod used for mounting the TLS was

ept at the same location between scans. The distance between theaser scanners (both ATLS and TLS) and the surveyed quaking aspenanopies ranged between 20 and 25 m.

.3. Data processing

All scan data were imported into the proprietary softwareackage Cyclone (Version 7.3, Leica Geosystems Inc., Heerbrugg,

en phenology and snow conditions during the outdoor experiment.

Switzerland). An open source software alternative for thisprocessing step could have been the use of LAStools. To determineplant canopy height (e.g., tree, graminoid, and herbs), the objectof interest was subset from the data and plant canopy height esti-mated based on the maximum and minimum z value (plant canopyheight = maximum z value − minimum z value). To determine DBH,the cross-sectional point cloud image was displayed in Cyclone andthe horizontal distance between the half-circle ends was taken withthe Cyclone distance tool. Trees that were partially obstructed byother trees in the sample plot were not included in the analysis.

To quantify the changes in canopy structural properties, wedetermined the number of canopy laser returns which has shownto provide valuable information about canopy structural proper-ties such as leaf area index or gap fraction (e.g., Clawges et al.,2007; Moorthy et al., 2008; Sanz-Cortiella et al., 2011b; Vierlinget al., 2012). Four quaking aspen canopies (hereafter referred toas canopy 1, canopy 2, canopy 3, and canopy 4) and one Ponderosapine canopy were visually isolated in the Cyclone software environ-ment and the point cloud was exported as a .txt file and importedinto R 2.14.1 to determine the number of canopy laser returns.

Increasing beam divergence results in a lower laser point cloudresolution (Van Gnechten et al., 2008). Hence, the number of canopylaser returns between ATLS and TLS are inherently different due tothe smaller beam divergence of the TLS (0.15 mrad) when com-pared to the ATLS (1.7 mrad). To thus allow a visual comparisonbetween ATLS and TLS mapped canopy structural changes, thereturn percentage of ATLS and TLS returns was calculated as fol-lows:

Return percentage = Nday x

NOctober 12× 100 (6)

where Nday x are the number of hits on day x of the experiment andNOctober 12 are the number of hits at the beginning of the experimenton October 12, 2012. Besides the number of hits, tree height for thequaking aspen from both the ATLS and TLS dataset was calculated

J.U.H. Eitel et al. / Agricultural and Forest Meteorology 180 (2013) 86– 96 91

Fig. 3. (A) Digital image of study site and autonomously operating terrestrial laserscanner (ATLS) and terrestrial laser scanner (TLS) scanning positions, (B) ATLS scano

bo

iwC( .Tta

ground, causing a mixed edge return. Laser return intensity alsomeasured by the ATLS could be used to exclude mixed edge returns

f study site and (C) TLS scan of study site.

ased on the maximum and minimum z value associated with eachf the given tree point clouds acquired October 12, 2012.

Meteorological measurements during the experiment, includ-ng mean air temperature, precipitation, snow depth, and mean

ind speed, were taken from the nearest Natural Resourcesonservation Service (NRCS) SNOwpack TELemetry (SNOTEL) sitehttp://www.wcc.nrcs.usda.gov/nwcc/site?sitenum=319&state=id)o visually monitor the site conditions during the experiment, a

ime-lapse camera (Wingscapes, Inc., Alabaster, AL) was mountedt the site.

Fig. 4. Digital photograph of Engelmann spruce (Picea engelmannii) trees and asso-ciated point cloud acquired by the autonomously operating terrestrial laser scanner(ATLS).

2.4. Statistical analysis

The manually measured (observed) plant canopy heights andDBH (dependent variables) were related to ATLS derived esti-mates of plant canopy heights and DBH (independent variables),respectively, in the open-source software package R 2.14.1. For theresultant model, the root mean square error (RMSE), the coefficientof determination (r2), slope and intercept were calculated. A RMSEvalue of 0.0 and a r2 of 1.0 indicated high precision and a slope of1.0 and intercept of 0.0 indicated high accuracy. A simple linearregression model was fit between ATLS and TLS measured canopylaser returns and tree height.

3. Results and discussion

3.1. ATLS basic test results

The manually measured tree canopy heights of the fiveEngelmann spruce trees were 88.9, 78.74, 70.49, 53.97, and28.58 cm (Fig. 4). The manually measured tree canopy heightswere in good agreement with ATLS derived tree canopy heights(r2 = 1.00, RMSE = 0.96 cm, slope = 0.98, intercept = 1.43) (Fig. 5).However, the agreement between manually measured tree canopyheight and ATLS derived tree canopy height decreased (r2 = 0.99,RMSE = 2.63 cm, slope = 1.08, intercept = 1.30) after adding an object1 m behind the trees. This finding is in broad agreement with pre-vious studies (e.g., Rosell et al., 2009; Hebert and Krotkov, 1992).For example, Rosell et al. (2009) found that laser rangefinders mayprovide intermediate distances between two plant objects insteadof the distance of the first plant object the laser pulse encounters.This is explained by the fact that edges of a foreground object mightsplit a single laser pulse so that it hits a second object in the back-

from the analysis by using range dependent threshold values (Eitelet al., 2010, 2011a; Balduzzi et al., 2011). Mixed edge returns

92 J.U.H. Eitel et al. / Agricultural and Forest Meteorology 180 (2013) 86– 96

Fig. 5. Relationship between observed (manually measured) and predicted(e

gt

fcgRmAahFHR

F(a

autonomously operating terrestrial laser scanner derived) Engelmann spruce (Piceangelmannii) canopy height.

enerally show to have considerably lower return intensity valueshan laser returns from a single object.

The manually measured canopy height of the graminoid rangedrom 4.26 to 43.69 cm. For the herb, the manually measuredanopy height ranged from 3.26 to 24.13. The ATLS derivedraminoid and herb canopy height strongly agreed (r2 = 0.98,MSE = 1.56 cm, slope = 1.04, intercept = −2.22) with the manuallyeasured canopy heights (Fig. 6). It is interesting to note that theTLS derived canopy height estimates for the graminoid were lessccurate than the ATLS derived canopy height estimates for theerb. This is likely explained by the fact that the leaf plate of theeathered Reed Grass is narrower than the leaf plate of the Daylily.ence, the likelihood of a laser pulse hitting the top of the Feathered

eed Grass is lower than in the case of the Daylily.

ig. 6. Relationship between observed (manually measured) and predictedautonomously operating terrestrial laser scanner derived) graminoid (Calamgrostis)nd herb (Hemerocallis) canopy height.

Fig. 7. Relationship between observed (manually measured) and predicted(autonomously operating terrestrial laser scanner derived) diameter at breast height(dbh) of Ponderosa Pine (P. ponderosa) and Douglas Fir (P. menziesii) trees.

3.2. ATLS field test results

The DBH of trees surveyed ranged from 24.1 cm to 100.7 cm.There was a strong agreement (r2 = 0.99, RMSE = 2.24 cm,slope = 0.99, intercept = 0.45) between manually measured DBHand ATLS derived DBH (Fig. 7). The precision of ATLS derivedDBH measurements is comparable with the precision obtainedwith TLS in earlier studies (Hopkinson et al., 2004; Watt andDonoghue, 2005). For instance, Watt and Donoghue (2005)showed a strong linear relationship between measured and TLSderived DBH. Combining DBH with other biometric measure-ments potentially derivable from ATLS data, such as tree taper,branching frequency, stem density or canopy extent, could aid inbuilding and validating forest growth and carbon sequestrationmodels.

During the field deployment of the ATLS, temperature rangedbetween −9.1 and 21.1 ◦C, daily precipitation ranged from 0 to3.2 cm, and snowfall depth at the study site ranged from 0 to7.6 cm. During these wide ranges of weather conditions, the ATLScontinued to operate without any unplanned interruptions or mal-functions. Evidence from time-lapse camera images and periodicvisits to the study site during snowfall events revealed no signif-icant snow accumulation on the ATLS likely due to the constantmovement (in both azimuth and zenith angle direction) and inter-nal heating of the laser unit. Further, visual examinations of theoptics during periodic visits did not show any condensation, dust,insects, and/or bird deposit on the optics. However, during rainfall,rain droplets accumulated on the optics.

For quaking aspen canopies, the TLS derived canopy heightstrongly agreed with the ATLS derived canopy height (r2 = 0.99,RMSE = 0.08 m, intercept = −0.38, slope = 1.01) (Fig. 8). The returnpercent of both ATLS and TLS showed to overall decrease through-out the experiment (Fig. 9) confirming visual observations ofleaf-drop (Fig. 2). In contrast, the return percent for both ATLSand TLS returns remained constant at around 100% for the ever-green Ponderosa Pine. The increase in both ATLS and TLS returnpercent between October 22 and 25 can be explained by the snow-

fall event during this time period that lead to snow accumulationon the quaking aspen branches (Fig. 2C) and thus the increase inreturn percentage. The return percentage of both ATLS and TLSfor Ponderosa pine did not increase during the snowfall event,

J.U.H. Eitel et al. / Agricultural and Forest Meteorology 180 (2013) 86– 96 93

Fao

ps

2Wmtittwbsp

Fa(

ig. 8. Relationship between terrestrial laser scanner (TLS) derived canopy heightnd autonomously operating terrestrial laser scanner (ATLS) derived canopy heightf quaking aspen (Populus tremuloides).

ossibly due to considerably fewer branch surfaces that allowednow to accumulate.

Other increases in ATLS return percent, for example on October0 and October 29, could be explained by wind or rain events.indy conditions might lead to the movement of leaves and thusay increase the likelihood that a given leaf area is scanned several

imes. During strong rains, such as on October 29 where precip-tation was highest during the experiment, raindrops might haveriggered additional laser returns. However, on October 17, duringhe course of another strong storm event with high rainfall andind speeds, the ATLS recorded fewer laser returns, most likely

ecause of the aspen leaf drop that occurred as a result of thistorm. When evaluating the returns from the coniferous ponderosaine trees, only the snow storm appeared to affect the number of

ig. 9. Changes in both autonomously operating terrestrial laser scanner (ATLS)nd terrestrial laser scanner (TLS) return percentage measured for a quaking aspenPopulus tremuloides) canopy and Ponderosa Pine (P. ponderosa) canopy.

Fig. 10. Relationship between autonomously operating terrestrial laser scanner(ATLS) and terrestrial laser scanner (TLS) canopy laser returns of equal targets.

recorded laser returns, most likely because of snow that had accu-mulated on the trees themselves. Hence, ATLS data acquired duringstrong wind-, rain-, or snowfall events should be excluded fromanalysis of structural change.

The TLS return percentages for the aspen show to be in generallower than the ATLS return percentages for the aspen (Fig. 9). Thisdiscrepancy can be explained by the differences in beam divergencebetween the ATLS (1.7 mrad) and the TLS (0.15 mrad) that affectsthe sensitivity of the laser instrument to structural changes in par-ticular at fine spatial scales. Overall, the number of ATLS canopylaser returns showed to correlate strongly with TLS canopy laserreturns (r2 = 0.86, RMSE = 265 laser return pts) (Fig. 10). The num-ber of TLS laser returns revealed to be more than double the numberof ATLS laser returns. This again can likely be explained by thesmaller beam divergence of the TLS (0.15 mrad) compared to theATLS (1.7 mrad).

3.3. Limitations and future research needs

While the above results illustrate the potential of ATLS to mon-itor and quantify 3-D ecosystem dynamics, it is important to notethat such results might vary depending on the distance betweenthe ATLS and the surveyed object (Prokop, 2008; Schaffhauseret al., 2008). For example, Schaffhauser et al. (2008) showed thatthe accuracy of TLS derived snow depth estimates decreased withincreasing distance between the TLS and the survey point due tothe widening of the laser beam diameter with distance and resul-tant decrease in positional accuracy of the surveyed point (VanGnechten et al., 2008). Further, the accuracy of the results might beaffected by weather conditions (Van Gnechten et al., 2008; Prokop,2008; Schaffhauser et al., 2008). The distance measurements oftime-of-flight laser sensors are affected by atmospheric conditionsincluding temperature, atmospheric pressure, and relative humid-ity. These atmospheric variables affect the refraction index of theatmosphere and thus the travel time of the laser pulse and thereofderived distance measurements (Van Gnechten et al., 2008). Also,poor weather conditions such as snowfall, rain, or fog might affectthe reliability of the TLS derived metrics (Prokop, 2008), while

measurements taken in windy conditions may influence vegeta-tion 3-D measurements because of vegetation movement duringthe scans. Moreover, the measurement set-up for the ATLS (e.g.,height above ground, location relative to objects of interest) should

9 d Forest Meteorology 180 (2013) 86– 96

besce

hat2cesitbtoicb(Eattmudw

sctoorTdwottTwtste9lmorcftoto

vpdev

Fig. 11. Changes in autonomously operating terrestrial laser scanner (ATLS) return

4 J.U.H. Eitel et al. / Agricultural an

e carefully considered before measurements are conducted tonsure unoccluded views on the objects of interest. A detailedtudy on measurement set-up effects on the quality of TLS derivedanopy structural information was conducted by Van der Zandet al. (2006) and the interested reader is referred to this work.

The wider beam divergence of the ATLS (1.7 mrad) describedere might be problematic when monitoring small objects suchs leaves at longer range distances. In our field study, the dis-ance from the ATLS to the plant canopies ranged between 20 and5 m and the results show that the ATLS allowed canopy structuralhanges to be quantified. Similarly, ATLS data provided reliablestimates of DBH of trees that were up to 26 m away from thecan location. However, depending on the size of the object ofnterest, the maximum distance between the object of interest andhe ATLS that allows collecting useful laser data may vary. Hence,efore employing the ATLS at a study site, users are encouragedo test the maximum distance between the ATLS and the objectf interest that ensures useful 3-D data. Users that are interestedn employing an ATLS but require a narrower beam divergenceould consider using a different laser distance tool (with a smalleream divergence) atop the pan-tilt unit such as the FLS-CH10Dimetix AG, Herisau, Switzerland) or the optoNCDT ILR (Micro-psilon Messtechnik GmbH & Co. KG, Ortenburg, Germany) which,t 10 m, have a laser spot diameter of 8 mm and 11 mm, respec-ively. Indeed, before assembling an ATLS, users are encouragedo think about system requirements, such as beam divergence,

easurement range, or laser wavelength, desirable for a partic-lar application. The design of an ATLS is not limited to the laseristance tool used in this study but rather could be easily modifiedith a different laser distance tool that fits particular user needs.

The accuracy of the ATLS metrics might also be affected by thecan resolution. For example, it is more likely that the tip of therown of an Engelmann spruce tree will be surveyed the higherhe vertical and horizontal scan resolution. Similarly, estimates ofther ecosystem dynamics might improve with increasing scan res-lution. For example, Hancock et al. (2008) used TLS data to captureill morphology on freshly mined soil. Based on approximately 25LS sample points per m2 they found that small rills were not wellefined. The authors concluded that a greater density of pointsould have been needed to accurately capture rill morphology

f small rills. However, with increasing scan resolution, the scanime will increase. For example, assuming a scan area defined byhe following pan-tilt limits (Panmax = 250◦, Panmin = 0◦, Tiltmax = 5◦,iltmin = −35◦), using Eq. (1), one can calculate that the scan timeould more than double from 1:02:05 (hours:minutes:seconds) if

he scan resolution is 2◦ to 2:46:40 at a scan resolution of 1◦. Thelow scan time of the ATLS described in this paper may be a limi-ation when compared to commercially available TLS systems. Forxample, scanning the quaking aspen stand with the ATLS required

h and 17 min compared to 1 min and 15 s with the TLS at simi-ar scan resolutions. Hence, a single ATLS would not be suitable for

onitoring highly dynamic 3-D ecosystem processes that changen a minute or hourly basis. Rather, its strength lies in quantifyingepeated, high temporal resolution changes in 3-D ecosystem pro-esses that can be tracked using repeated scans at a data collectionrequency of 1–3 scans per day (frequency depends on scan resolu-ion and area of interest). Refinements of the CRBasic program thatperates the ATLS could help to increase the scan rate of the ATLS,hough it was beyond the scope of this proof-of-concept study toptimize the ATLS scan rate.

Finally, though the ATLS showed to operate reliably under a wideariety of weather conditions, additional field testing for extended

eriods of time and under a wider range of environmental con-itions not tested here (e.g., high humidity, strong rains for anxtended period of time, dust storms, high diurnal air temperatureariations, etc.) will help to further improve our understanding of

percentage measured during the leaf drop-off of a quaking aspen canopy. Daily pre-cipitation, snow depth, and mean wind speed are shown to aid interpretation ofstructural dynamics occurring in the ATLS return percentage.

what environmental conditions might adversely affect the reliableoperation of the ATLS.

3.4. Opportunities

The high temporal resolution of the ATLS data can provideinsights about 3-D ecosystem dynamical processes that could notbe provided by lower temporal resolution TLS data. This point isillustrated by canopy structural changes of quaking aspen quanti-fied by both ATLS and TLS in this study (Fig. 11). Though both ATLSand TLS data capture the leaf drop-off, the lower temporal reso-lution TLS data does not allow determining the time-period whenmost of the leaf drop-off occurred and thus the key event associatedwith canopy structural changes in this study. In contrast, the hightemporal resolution ATLS dataset shows that the largest leaf drop-off occurred between October 16 and 18 which coincided with highwind speeds (up to 14.3 km/h) that triggered the leaf drop-off.

The relatively light weight (<4 kg vs. >9 kg for most TLS systems)and small size of the ATLS could further allow selecting scan loca-tions that would be difficult to access or reach with traditionalTLS. For example, the Leica ScanStation2 weighs 25 kg (includingtransportation case and accessories such as cables and tribrach)which makes the transportation of the device to some areas diffi-cult, and can add large shipping costs to field sampling campaigns.Also, the light weight of the ATLS allows more flexible mountingoptions when compared to a TLS. For example, the ATLS couldbe mounted onto a tall tripod or tower system where it is gen-erally difficult to mount a TLS. This could be a great advantage indensely vegetated study areas where occlusion often complicatesTLS surveys, and would allow the ATLS measurements to be inte-grated into SNOTEL sites or optical measurement schemes locatedat flux towers (e.g., Gamon et al., 2006). Such use would also beadvantageous for using ATLS measurements to validate aircraft-or satellite-collected RADAR or LiDAR measurements of vegetationstructure.

Another key advantage of the ATLS is that it does notrequire the establishment of a reference target network gen-

erally necessary to register multi-temporal TLS scans to acommon coordinate system. The origin of point cloud coordinatesystems (x = 0, y = 0, z = 0) created by both ATLS and TLS is the cen-ter of the instrument. Temporally subsequent ATLS scans have the

d Fore

sEIcstcstdrHts

>(aa∼aprbaama

dctiBaris

4

it$sfrqwvAtrsasrtietgm

J.U.H. Eitel et al. / Agricultural an

ame coordinate system since the real world coordinates (Lat, Long,levation) of the ATLS instrument center does not change over time.n contrast, the real world coordinates of the instruments centerommonly change between temporally subsequent TLS data acqui-itions. Thus, temporally invariant targets are generally used inranslation and rotation algorithms to bring all scans into a commonoordinate system. The establishment of such a reference targetystem can be time consuming and costly when artificial referenceargets are employed (e.g., Eitel et al., 2011b) and may be prone toamage of removal. Alternatively, natural targets have been used toegister multi-temporal point clouds (Henning and Radtke, 2006).owever, it may prove difficult to find suitable reference targets

hat are stable over time – especially when monitoring dynamicystems (Avian et al., 2009).

Compared to traditional TLS (with entry level systems costing$40,000 as of 2012), the overall cost of the ATLS is relatively low$11,841 USD as of 2012; Table 1). One of the major costs associ-ted with the system is the laser rangefinder ($5575). In this study,

laser rangefinder was chosen that has a measurement range of300 m for natural surfaces such as plants or soil. For studies thatttempt to quantify and monitor 3-D ecosystem changes in studylots with a diameter of considerably less than 300 m radius, a laserangefinder with a lower distance range and thus lower price coulde sufficient. However, it is important to consider that plant leavesbsorb strongly in the red spectral region (reflectance ranges atround 1–5% depending on the plant chlorophyll content) whichight considerably lower the quoted maximum distance range of

laser rangefinder if the ATLS is used for vegetation monitoring.In addition to monitoring and quantifying 3-D ecosystem

ynamics, ATLS might also be useful to monitor and quantify bio-hemical changes of natural surfaces. Recent research has shownhat the laser return intensity of a green scanning laser can providenformation about foliar biochemistry (Eitel et al., 2010, 2011a).esides distance, the ATLS described in this paper also measuresnd records the laser return intensity of its NIR laser. Furtheresearch is needed to test the suitability of the NIR laser returnntensity to monitor and quantify biochemical variability of naturalurfaces.

. Conclusion

We described the design of a low-cost, autonomously operat-ng terrestrial laser scanner or ATLS. Compared to traditional TLS,he cost of the system is relatively low (ATLS: <$12,000 USD vs. >40,000 USD for entry level TLS). Though the scanning rate of theensor is 1 pulse per 1.5 s limiting its use for studies that requireast scan rates, the strength of the ATLS is its ability to acquireepeated, high temporal resolution TLS data (1–3 scans per day, fre-uency depends on scan resolution and area of interest). The ATLSas tested under both controlled laboratory conditions, and highly

ariable field conditions. The laboratory findings revealed that theTLS provides useful 3-D data of plant structural changes. Testing

he ATLS under field conditions revealed that the ATLS provideseliable estimates of DBH and information about dynamic canopytructural changes during leaf drop-off in quaking aspen. The ATLSlso operated reliably under harsh weather conditions, includingnow, rain, fog and wide temperature ranges (−9.1 to 21.1 ◦C). Theesults of this study are encouraging and show the potential of ATLSo monitor and quantify 3-D ecosystem dynamics. Such informations of great importance to further our understanding of inher-

ntly dynamic ecosystem processes. Further research is neededo better understand the effects of scan resolution, beam diver-ence, and atmospheric conditions on the accuracy of ATLS derivedetrics.

st Meteorology 180 (2013) 86– 96 95

Acknowledgements

We would like to thank Mrs. Donna and the late Dr. Her-ald Nokes and Family for granting us access to the Herald NokesExperimental Forest. We would also like to thank Bridget Halland Patrick O’Keeffe for their help with the CRBasic program-ming, David Brown and Sanford Eigenbrode for their support inthe development of the sensor package, Josh Jones and Ed Quer-feld from Micro-Epsilon who provided valuable technical insightsconcerning the laser rangefinder used in this study, Martial Hebertfor reviewing the mixed edge description in the introduction, andtwo anonymous reviewers for their helpful comments to improveearlier versions of this manuscript. This work was supported byUSDA-NIFA Award Nos. 2011-67003-3034 and 2011-68002-30191,and Research Grant Award No. L08AC14585 by the Bureau of LandManagement. We also acknowledge funding support from NASATerrestrial Ecology grant NNX12AK83G. Funding to acquire the TLSused in this study was provided by the University of Idaho, IdahoNSF EPSCoR, and by the National Science Foundation under AwardNumber EPS-0814387. Use of trade names does not constitute anofficial endorsement by the University of Idaho.

References

Adler, D., Murdoch, D., 2011. rgl: 3-D Visualization Device System (OpenGL). RPackage version 0.92.798. http://CRAN.R-project

Arora, V., 2002. Modeling vegetation as a dynamic component in soil-vegetation-atmosphere transfer schemes and hydrological models. Rev. Geophys. 40, 3-1–3-26.

Avian, M., Kellerer-Pirklbauer, A., Bauer, A., 2009. LiDAR for monitoring mass move-ments in permafrost environments at the cirque Hinteres Langtal, Austria,between 2000 and 2008. Nat. Hazards Earth Syst. Sci. 9, 1087–1094.

Balduzzi, M.a.F., Van der Zande, D., Stuckens, J., Verstraeten, W.W., Coppin, P., 2011.The properties of terrestrial laser system intensity for measuring leaf geome-tries: a case study with Conference Pear trees (Pyrus communis). Sensors 11,1657–1681.

Clark, R.N., 1999. Spectroscopy of rocks and minerals, and principles of spectroscopy.In: Rencz, A. (Ed.), Manual of Remote Sensing. Remote Sensing for the EarthSciences, vol. 3. John Wiley and Sons, New York, pp. 3–58 (Chapter 1).

Clawges, R., Vierling, L., Calhoon, M., Toomey, M., 2007. Use of a ground-basedscanning lidar for estimation of biophysical properties of western larch (Larixoccidentalis). Int. J. Remote Sens. 28, 4331–4344.

Conner, J., 2010. LViz: 3D LiDAR Visualization Tool, Developed with Funding fromthe NSF GEON Project.

Egli, L., Jonas, T., Grünewald, T., Schirmer, M., Burlando, P., 2012. Dynamics of snowablation in a small Alpine catchment observed by repeated terrestrial laser scans.Hydrol. Process. 26, 1574–1585.

Eitel, J.U.H., Long, D.S., Gessler, P.E., Hunt, E.R., Brown, D.J., 2009. Sensitivity ofground-based remote sensing estimates of wheat chlorophyll content to varia-tion in soil reflectance. Soil Sci. Soc. Am. J. 73, 1715–1723.

Eitel, J.U.H., Vierling, L.A., Long, D.S., 2010. Simultaneous measurements of plantstructure and chlorophyll content in broadleaf saplings with a terrestrial laserscanner. Remote Sens. Environ. 114, 2229–2237.

Eitel, J.U.H., Vierling, L.A., Long, D.S., Hunt, E.R., 2011a. Early season remote sensingof wheat nitrogen status using a green scanning laser. Agric. Forest Meteorol.151, 1338–1345.

Eitel, J.U.H, Williams, C.J., Vierling, L.a., Al-Hamdan, O.Z., Pierson, F.B., 2011b. Suit-ability of terrestrial laser scanning for studying surface roughness effects onconcentrated flow erosion processes in rangelands. Catena 87, 398–407.

Gamon, J., Rahman, a., Dungan, J., Schildhauer, M., Huemmrich, K., 2006. Spectralnetwork (SpecNet)—what is it and why do we need it? Remote Sens. Environ.103, 227–235.

Gausman, H.W., 1985. Plant Leaf Optical Properties in Visible and Near-infrared Light(Graduate Studies). Texas Tech University Press.

Gutmann, E., 2010. Continuous alpine snow depth mapping by laser rangefinderthrough a winter season. In: AGU Fall Meeting Abstracts 2010.

Gutmann, E., Larson, K.M., Williams, M.W., Nievinski, F.G., Zavorotny, V., 2011. Snowmeasurement by GPS interferometric reflectometry: an evaluation at NiwotRidge, Colorado. Hydrol. Process., 8329, http://dx.doi.org/10.1002/hy.

Hancock, G., Crawter, D., Fityus, S., 2008. The measurement and modelling of rillerosion at angle of repose slopes in mine spoil. Earth Surf. 1020, 1006–1020.

Haubrock, S.-N., Kuhnert, M., Chabrillat, S., Güntner, A., Kaufmann, H., 2009. Spa-tiotemporal variations of soil surface roughness from in-situ laser scanning.

Catena 79, 128–139.

Hebert, M., Krotkov, E., 1992. 3D measurements from imaging laser radars: howgood are they? Image Vision Comput. 10, 170–178.

Henning, J., Radtke, P.J., 2006. Ground-based laser imaging for assessing threedimensional forest canopy structure. Photogramm. Eng. Remote 72, 1349–1358.

9 d Fore

H

I

I

J

K

K

L

L

M

M

M

P

PP

R

R

R

R

S

S

Williams, R., Brasington, J., Vericat, D., Hicks, M., Labrosse, F., Neal, M., 2011.

6 J.U.H. Eitel et al. / Agricultural an

opkinson, C., Chasmer, L., Young-Pow, C., Treitz, P., 2004. Assessing forest metricswith a ground-based scanning lidar. J. Forest 583, 573–583.

senburg, M., 2007–2012. LAStools – Efficient Tools for LiDAR Processing.http://lastools.org

ves, A.R., Carpenter, S.R., 2007. Stability and diversity of ecosystems. Science 317,58–62.

ester, W., Klik, A., 2005. Soil surface roughness measurement—methods, applica-bility, and surface representation. Catena 64, 174–192.

eightley, K.E., Bawden, G.W., 2010. 3D volumetric modeling of grapevine biomassusing Tripod LiDAR. Comput. Electron. Agric. 74, 305–312.

im, H., Arrowsmith, J.R., Crosby, C.J., Jaeger-Frank, E., Nandigam, V., Memon, A.,Conncer, J., Badden, S.B., Baru, C.n.d, 2006. An efficient implementation of a localbinning algorithm for digital elevation model generation of LiDAR/ALSM dataset.Eos Trans. AGU 87 (52) (Fall Meet. Suppl., Abstract G53C-0921).

efsky, M.A., Cohen, W.B., Parker, G.G., Harding, D.J., 2002. Lidar remote sensing forecosystem studies. Bioscience 52, 19–30.

im, M., Petley, D., Rosser, N., 2005. Combined digital photogrammetry and time-of-flight laser scanning for monitoring cliff evolution. Photogramm. Rec. 20,109–129.

ilan, D.J., Heritage, G.L., Hetherington, D., 2007. Application of a 3D laser scannerin the assessment of erosion and deposition volumes and channel change in aproglacial river. Earth Surf. Process. Landforms 32, 1657–1674.

oorthy, I., Miller, J., Hu, B., Chen, J., Li, Q., 2008. Retrieving crown leaf area indexfrom an individual tree using ground-based lidar data. Can. J. Remote Sens. 34,320–332.

oorthy, I., Miller, J.R., Berni, J.A.J., Zarco-Tejada, P., Hu, B., Chen, J., 2011. Field char-acterization of olive (Olea europaea L.) tree crown architecture using terrestriallaser scanning data. Agric. Forest Meteorol. 151, 204–214.

erroy, R.L., Bookhagen, B., Asner, G.P., Chadwick, O.a., 2010. Comparison of gullyerosion estimates using airborne and ground-based LiDAR on Santa Cruz Island,California. Geomorphology 118, 288–300.

arker, G.G., Harding, D.J., Berger, M.L., 2004. J. Appl. Ecol. 41, 755–767.rokop, A., 2008. Assessing the applicability of terrestrial laser scanning for spatial

snow depth measurements. Cold Reg. Sci. Technol. 54, 155–163. Development Core Team, 2011. R: A Language and Environment for Statistical

Computing. R Foundation for Statistical Computing, Vienna, Austria, ISBN 3-900051-07-0 http://www.R-project.org/

oncat, A., Székely, B., Molnár, G., Zámolyi, A., Pocsai, A., Ressl, C., 2011. High-resolution terrestrial laser scanning of temporary small-scale terraces in adynamic landscape—a pilot study at the Doren landslide. Geophys. Res. Abstr.13, 13847.

osell, J.R., Llorens, J., Sanz, R., Arnó, J., Ribes-Dasi, M., Masip, J., Escolà, A., Camp, F.,Solanelles, F., Gràcia, F., Gil, E., Val, L., Planas, S., Palacín, J., 2009. Obtaining thethree-dimensional structure of tree orchards from remote 2D terrestrial LIDARscanning. Agric. Forest Meteorol. 149, 1505–1515.

yu, Y., Verfaillie, J., Macfarlane, C., Kobayashi, H., Sonnentag, O., Vargas, R., Ma, S.,Baldocchi, D.D., 2012. Continuous observation of tree leaf area index at ecosys-tem scale using upward-pointing digital cameras. Remote Sens. Environ. 126,116–125.

ankey, J.B., Eitel, J.U.H., Glenn, N.F., Germino, M.J., Vierling, L.A., 2011. Quantify-ing relationships of burning, roughness, and potential dust emission with laser

altimetry of soil surfaces at submeter scales. Geomorphology 135, 181–190.

anz-Cortiella, R., Llorens-Calveras, J., Rosell-Polo, J.R., Gregorio-Lopez, E., Palacin-Roca, J., 2011a. Characterisation of the LMS200 laser beam under the influenceof blockage surfaces. Influence on 3D scanning of tree orchards. Sensors 11,2751–2772.

st Meteorology 180 (2013) 86– 96

Sanz-Cortiella, R., Llorens-Calveras, J., Escolà, A., Arnó-Satorra, J., Ribes-Dasi, M.,Masip-Vilalta, J., Camp, F., Gràcia-Aguilá, F., Solanelles-Batlle, F., Planas-DeMartí,S., Pallejà-Cabré, T., Palacin-Roca, J., Gregorio-Lopez, E., Del-Moral-Martínez, I.,Rosell-Polo, J.R., 2011b. Innovative LIDAR 3D Dynamic Measurement System toestimate fruit-tree leaf area. Sensors 11, 5769–5791.

Scanlon, T.M., Caylor, K.K., Manfreda, S., Levin, S.a., Rodriguez-Iturbe, I., 2005.Dynamic response of grass cover to rainfall variability: implications for the func-tion and persistence of savanna ecosystems. Adv. Water Resour. 28, 291–302.

Schaffhauser, A., Adams, M., Fromm, R., Jörg, P., Luzi, G., Noferini, L., Sailer, R., 2008.Remote sensing based retrieval of snow cover properties. Cold Reg. Sci. Technol.54, 164–175.

Schurr, U., Walter, A., Rascher, U., 2006. Functional dynamics of plant growth andphotosynthesis—from steady-state to dynamics—from homogeneity to hetero-geneity. Plant Cell Environ. 29, 340–352.

Schwalbe, E., Maas, H., Dietrich, R., Ewert, H., 2008. Glacier velocity determina-tion from multi temporal terrestrial long range laser scanner point clouds.In: International Archives of the Photogrammetry, Remote Sensing and SpatialInformation Sciences, vol. XXXVII. Part B5, Beijing, pp. 457–462.

Shugart Jr., H.H., Crow, T.R., Hett, J.M., 1973. Forest succession models: a rationaleand methodology for modeling forest succession over large regions. Forest Sci.19, 203–212.

Staley, D.M., Wasklewicz, T.A., Coe, T.A., Kean, J.W., McCoy, S.W., Tucker, G.E., 2011.Observations of debris flows at Chalk Cliffs, CO, USA: part 2, changes in sur-face morphometry from terrestrial laser scanning in the summer of 2009. In:5th International Conference on Debris-Flow Hazards Mitigation: Mechanics,Prediction and Assessment, Padua, Italy, June 14–17, pp. 759–768.

Starek, M., Mitasova, H., Hardin, E., 2011. Modeling and analysis of landscape evo-lution using airborne, terrestrial, and laboratory laser scanning. Geosphere 7,1340–1356.

Treuhaft, R.N., Cloude, S.R., 1999. The structure of oriented vegetation from polari-metric interferometry. IEEE Trans. Geosci. Remote Sens. 37, 2620–2624.

Tuley, J., Vandapel, N., Hebert, M., 2004. Analysis and Removal of Artifacts in 3-DLADAR Data. Robotics Institute (Paper 338).

Van der Zande, D., Hoet, W., Jonckheere, I., Vanaardt, J., Coppin, P., 2006. Influenceof measurement set-up of ground-based LiDAR for derivation of tree structure.Agric. Forest Meteorol. 141, 147–160.

Van Gnechten, B., Caner, H., Heine, E., García, J.L.L., Poelman, R., Quintero, M.S., 2008.Theory and Practice on Terrestrial Laser Scanning., pp. 14–35.

Vierling, K.T., Vierling, L.A., Gould, W.A., Martinuzzi, S., Clawges, R.M., 2008. Lidar:shedding new light on habitat characterization and modeling. Front. Ecol. Envi-ron. 6, 90–98.

Vierling, L.A., Yu, Y., Eitel, J.U.H., Oldow, J.S., 2012. Shrub characterization using ter-restrial laser scanning and implications for airborne LiDAR assessment. Can. J.Remote Sens. 38, 1–14.

Watt, P.J., Donoghue, D.N.M., 2005. Measuring forest structure with terrestrial laserscanning. Int. J. Remote Sens. 26, 1437–1446.

Wenske, D., Jen, C.-H., Böse, M., Lin, J.-C., 2012. Assessment of sediment deliveryfrom successive erosion on stream-coupled hillslopes via a time series of topo-graphic surveys in the central high mountain range of Taiwan. Quatern. Int. 263,14–25.

Case study: monitoring braided river change using terrestrial laser scan-ning and optical bathymetric mapping. In: Smith, M.J., Paron, P., Griffiths, J.S.(Eds.), Geomorphological Mapping: Methods and Applications. Elsevier, Ams-terdam/Lubbock, Texas, USA, pp. 507–532.