Remote Sensing of Environment 251 (2020) 112102

Contents lists available at ScienceDirect

Remote Sensing of Environment

journal homepage: www.elsevier.com/locate/rse

Review

Terrestrial laser scanning in forest ecology: Expanding the horizon T

a,⁎ b c d e
Kim Calders , Jennifer Adams , John Armston , Harm Bartholomeus , Sebastien Bauwens ,
Lisa Patrick Bentleyf, Jerome Chaveg, F. Mark Dansonh, Miro Demola,i, Mathias Disneyj,k,
Rachel Gaultonl, Sruthi M. Krishna Moorthya, Shaun R. Levickm, Ninni Saarinenn,o,
Crystal Schaafp, Atticus Stovallq, Louise Terryna, Phil Wilkesj,k, Hans Verbeecka
a
CAVElab - Computational & Applied Vegetation Ecology, Department of Environment, Ghent University, Belgium
b
European Space Agency, ESA-ESRIN, 00044 Frascati, Italy
c
Department of Geographical Sciences, University of Maryland, College Park, MD 20742, USA
d
Laboratory of Geo-Information Science and Remote Sensing, Wageningen University & Research, Droevendaalsesteeg 3, 6708, PB, Wageningen, the Netherlands
e
Forest is Life, Gembloux Agro-Bio Tech, University of Liège, 5030 Gembloux, Belgium
f
Department of Biology, Sonoma State University, 1801 E. Cotati Ave., Rohnert Park, CA 94928, USA
g
Laboratoire Evolution et Diversité Biologique UMR 5174 CNRS, IRD, Université Paul Sabatier, 31062 Toulouse, France
h
School of Science, Engineering and Environment, University of Salford, Salford M5 4WT, UK
i
Centre of Excellence PLECO (Plants and Ecosystems), Biology Department, University of Antwerp, Universiteitsplein 1, B-2610 Wilrijk, Belgium
j
UCL Department of Geography, Gower Street, London WC1E 6BT, UK
k
NERC National Centre for Earth Observation (NCEO), UK
l
School of Natural and Environmental Sciences, Newcastle University, Newcastle Upon Tyne NE1 7RU, UK
m
CSIRO, PMB 44, Winnellie, 0822, NT, Australia
n
Department of Forest Sciences, University of Helsinki, P.O. Box 27, 00014 Helsinki, Finland
o
School of Forest Sciences, University of Eastern Finland, P.O. Box 111, 80101 Joensuu, Finland
p
School for the Environment, University of Massachusetts Boston, Boston, MA, USA
q
NASA Goddard Space Flight Center, 8800 Greenbelt Rd., Greenbelt, MD, United States

ARTICLE INFO ABSTRACT

Keywords: Terrestrial laser scanning (TLS) was introduced for basic forest measurements, such as tree height and diameter,
Terrestrial laser scanning in the early 2000s. Recent advances in sensor and algorithm development have allowed us to assess in situ 3D
Ground-based LiDAR forest structure explicitly and revolutionised the way we monitor and quantify ecosystem structure and function.
Forest ecology Here, we provide an interdisciplinary focus to explore current developments in TLS to measure and monitor
Forest plot measurement
forest structure. We argue that TLS data will play a critical role in understanding fundamental ecological
Tree structure
Remote sensing
questions about tree size and shape, allometric scaling, metabolic function and plasticity of form. Furthermore,
these new developments enable new applications such as radiative transfer modelling with realistic virtual
forests, monitoring of urban forests and larger scale ecosystem monitoring through long-range scanning. Finally,
we discuss upscaling of TLS data through data fusion with unmanned aerial vehicles, airborne and spaceborne
data, as well as the essential role of TLS in validation of spaceborne missions that monitor ecosystem structure.

1. Introduction understanding how terrestrial ecosystems are functioning and physi­

cally changing due to climate change (Calders et al., 2020; Verbeeck
Prior to the availability of laser scanning, explicit 3D forest structure et al., 2019).
was often represented qualitatively. These representations were often The potential of TLS for forest monitoring was first demonstrated in
only two dimensional such as the hand-drawn tropical forest tree ar­ published literature since the early 2000s (Hopkinson et al., 2004; Jupp
chetypes (Hallé et al., 1978; Specht, 1970). New developments in ter­ et al., 2009; Lovell et al., 2003; Strahler et al., 2008). Applications in­
restrial laser scanning (TLS) provide unprecedented three-dimensional itially focused on measuring traditional structural metrics that are used
in situ information of trees and forests (Malhi et al., 2018). This 3D in forestry such as height and diameter at breast height (DBH), but
information is argued to play a key role in monitoring and eventually evolved to whole-tree volumetric assessment to improve

⁎
Corresponding author.
E-mail address: [email protected] (K. Calders).

https://doi.org/10.1016/j.rse.2020.112102
Received 20 April 2020; Received in revised form 8 September 2020; Accepted 11 September 2020
Available online 25 September 2020
0034-4257/ © 2020 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/BY/4.0/).
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

estimates of aboveground biomass (Calders et al., 2015a; Gonzalez de range TOF, [c] long-range PS, [d] long-range TOF, and [e] mid-range
Tanago et al., 2018; Momo Takoudjou et al., 2018). Current applica­ dual wavelength, highlighting relevant forest ecology applications and
tions also include individual modelling of branch architecture (Lau avenues for sensor improvement in an operational context. Table 1
et al., 2018), habitat assessment (Ashcroft et al., 2014) or quantifying gives a non-exhaustive overview of commonly used commercial and
fuel loads (Chen et al., 2016; Hudak et al., 2009; Loudermilk et al., non-commercial TLS instruments in vegetation monitoring.
2009). Vertical profiles of forest structure from TLS (Palace et al., 2016) Short-range TOF TLS systems may provide an affordable option for
can characterise successional vegetation types (Cuni-Sanchez et al., high-quality TLS data, at the expense of limited range and/or dur­
2016) and show potential for global ecology and biodiversity studies ability. The canopy biomass lidar (CBL) is a non-commercial multiple-
(Valbuena et al., 2020) when combined with large footprint spaceborne return low-range (< 40 m) TLS built for durability (e.g. covered ro­
LiDAR missions, such as GEDI (Global Ecosystem Dynamics Investiga­ tating mirror and IP68 weatherproof) and fast acquisition time (Paynter
tion) (Marselis et al., 2018). et al., 2016, 2018, 2019). Due to range limitations, the CBL is ideal for
Here, we provide an interdisciplinary view on essential method quickly characterising short forests and the under- and mid-story of
development and technological advances in TLS in forest ecology. forests using a metaproperty approach (Paynter et al., 2018). The Leica
These new developments are set to revolutionise the way in which we BLK360 – a less durable (e.g. exposed rotating mirror and non-water­
observe and monitor changes in tree and forest structure and func­ proof) mid-range (60 m) TOF TLS is a lower-cost commercial option
tioning. We review and identify state-of-the-art methods for a range of that offers a cost-effective alternative for quantifying forest structure in
ecological applications, and reflect on their bottlenecks and current detail, with initial observations showing high-quality point clouds with
issues. We aim to set an agenda for increased uptake of TLS to support lower noise than PS TLS units in a similar price bracket (Disney et al.,
and improve our understanding and dynamic modelling of forest eco­ 2019). However, increased range and durability would be needed to
systems in a changing climate. Within this context, we first discuss enable use in taller stature forests and more challenging field condi­
critical sensor and algorithm development and their role in forest tions.
measurements. We then reflect on new opportunities that these in situ PS TLS instruments are generally cheaper, lighter, and capture high-
3D data are creating, including research on tree form and function, resolution data effective in low to mid complexity or leaf-off forests.
input for radiative transfer modelling, monitoring of urban trees and Due to their light weight (~5 kg), PS TLS are ideal in difficult terrain or
long-range scanning of large areas. Finally, we will discuss the up­ secluded field sites, where heavier scanners are more cumbersome
scaling of fundamental ecological understanding through TLS data and when traveling long distances on foot. Scan noise is highest in con­
links to large-area or global remote sensing products. iferous and dense broadleaf forests (Newnham et al., 2012) since PS
technology has greater ranging ambiguity when intercepting multiple
2. Advances in instrument technology and data processing objects within the footprint of a single laser beam. Filtering and mod­
elling algorithms specific to PS TLS focus on reducing ranging errors by
2.1. Advancing TLS technology enables sensor-specific applications excluding low return intensity and inconsistent range estimates by di­
rectly filtering outliers at the scan grid-level. Appropriate filtering of PS
While the basic premise for collecting 3D data is similar across TLS data enables accurate and precise biomass estimates of trunk, branch,
instruments, distinct ranging methods have emerged: time-of-flight and needles (~3–15% RMSE for whole-tree biomass) in coniferous trees
(TOF) and phase-shift (PS) sensors (Kahlmann et al., 2006; Wehr and (Stovall et al., 2017). In leaf-off hardwood forests, phase-shift sensors
Lohr, 1999). They differ primarily by a balance of cost and signal-to- can help develop non-destructive allometric models (Stovall et al.,
noise ratio (SNR). TOF sensors emit a discrete outgoing laser pulse, 2018), reducing uncertainty in calibration and validation of global
measuring the amount of time required to intercept an object and re­ aboveground biomass missions (Stovall and Shugart, 2018). Given the
turn to the scanner. Using the constant speed of light in vacuum low cost of PS instruments, noise reduction (e.g. more advanced grid
(299,792,458 m s−1) and the laser pulse travel time divided by two, filtering or post-processing k-nearest neighbor and outlier removal)
distance is estimated. PS sensors operate using a similar concept, but would be the main requirement to substantially improve the applic­
emit a continuous signal, modulating frequency and amplitude to ability of these instruments in more complex forests.
produce a unique outgoing signal. The incoming signal is compared Long-range (> 300 m) TOF TLS systems are widely recognized as
with the outgoing signal and out-of-phase regions are translated into the best option for measuring forest structure and tree architecture in a
range estimates. PS sensors tend to be quick, relatively inexpensive, range of forest types with the highest spatial detail (Newnham et al.,
lightweight, and have low beam divergence that produces extremely 2015). TLS instruments in this category, e.g. RIEGL VZ400(i) or Leica
high-resolution data, but rarely sense multiple returns, have lower SNR, ScanStation P40, are ideal for capturing forest structure for several
lower maximum range, and increased ranging artifacts compared to reasons. First, high-powered TOF instruments have superior SNR, pro­
TOF sensors (Newnham et al., 2012). The latter is especially true in viding more accurate return positions with less noise (Newnham et al.,
complex vegetation, where multiple beam interceptions occur, sub­ 2012). This, coupled with over 1 MHz pulse rate, the ability to detect
stantially increasing range ambiguity in single return systems. Despite multiple returns per laser pulse, and near-automatic registration, in­
TOF sensors generally being heavier and more expensive, their con­ creases measurement density and reduces occlusion (Calders et al.,
sistent, high-quality data they collect make them the current “gold- 2014). The tight manufacturer specifications in high-end instruments
standard” TLS (Fig. 1) for a range of vegetation applications (Newnham also improve instrument comparability (Calders et al., 2017) and global
et al., 2012). For all TLS instruments, the effective range of measure­ consistency in TLS products (Duncanson et al., 2019).
ments is limited by surface reflectance, which can be estimated directly The spectral information from TLS return intensity can be leveraged
from instrument specifications with the radar equation. For example, at to estimate biochemical properties and separate leaves from wood. Eitel
an approximate leaf reflectance of 20%, long-range TOF scanners can et al. (2010) and Zhu et al. (2015) used the intensity of a typical single-
reliably collect measurements at ~150 m distance, while other laser wavelength commercial TLS instrument to estimate biochemical para­
scanning technologies are effectively limited to a range of ~30–40 m. meters (chlorophyll and water content, respectively). However, this
Beam divergence affects the resolution of fine branches and leaves in approach can be challenging as this requires adjusting for incidence
the TLS data, so should be carefully considered, depending on forest angle and partial intercepted laser beams and needs further research in
type and the specific application of TLS data. more complex and larger canopies. An evaluation of the radiometric
Due to cost, speed of acquisition, weight, and SNR, specific sensors calibration of three same make and model (RIEGL VZ-400) scanners
and approaches may be ideal, depending on the application. In order showed that radiometric calibrations are instrument specific and that
from least to most expensive, we evaluate [a] short-range TOF, [b] mid- absolute bias is greater for high reflectance returns (Calders et al.,

2
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

Fig. 1. 3D complexity of a Simple Notophyll Vine Forest (Robson Creek, Australia) captured using a time-of-flight RIEGL VZ-400 instrument. Left panel: The colours
represent the distance from the scanner. Right panel: Derived plant area volume density as a function of canopy height derived for the same scan using (Calders et al.,
2015b). This estimated structural metric tells us how the volume of leaf and branch material is distributed with height in the canopy and its integral is the plant area
index.

Table 1
Examples of TLS instruments used to assess forest structure.
Major Instrument Short-range TOF + Mid-range TOF + Long-range PS + small Long-range TOF + Mid-range Dual
Categories large beam divergence medium beam beam divergence medium beam Wavelength + medium
divergence divergence + low noise beam divergence

Cost $ $ $$ $$$ $$$

Ideal Forest conditions +Sparse/simple +Sparse/simple + Leaf-off or + Best in tall/dense + Accessible, structurally
forests forests structurally simple forests simple forest stands
+ Remote areas + Remote areas forest stands
+ Remote areas
Optimal Forestry + Rapid assessment + Rapid assessment + Finely resolving + Finely resolving + Leaf-wood separation
Applications + Robust + Cost-effective forest small branches small branches + Biochemical properties
+ Cost-effective forest structural metrics + Potential for full + Improved vertical foliage
structural metrics waveform applications distribution
+ Potential for full
waveform applications
Example Instrument UMB CBL (SICK Lidar; Leica BLK360 FARO Focus3D X 330 RIEGL-VZ400i SALCA
non-commercial) (non-commercial)
Ranging method TOF TOF PS TOF TOF
# returns 1st + 2nd Single Single Multiple Full waveform
Wavelength [nm] 905 830 1550 1550 1545.4 & 1063.4
Maximum Range [m] 40 0.6 - 60 0.6 - 330 1.5 – 250 (high speed) 100 m
0.5 – 800 (long range)
Samples/sec 11,000 360,000 122,000- 976,000 42,000– 500,000 5,000
Beam Divergence [mrad] 15 0.4 0.19 0.35 0.56
Weight [kg] 3.9 1 5.2 9.7 17
Temperature range [deg C] -30 to 50 5 to 40 5 to 40 0 to 40 5 to 30
References Paynter et al. (2018, Disney et al. (2019) Liang et al. (2015); Bienert et al. (2018); Danson et al. (2018);
2016) Pyörälä et al. (2018) Tian et al. (2019) Schofield et al. (2016)

2017). Users are therefore recommended to carry out a radiometric function relationships through canopy metrics (Atkins et al., 2018a,
calibration before including reflectance information in TLS analysis 2018b; Parker et al., 2004), but the instrument has yet to be widely
(Calders et al., 2017; Hartzell et al., 2015). Including intensity-based adopted. Handheld mobile laser scanning (MLS) continuously acquires
features through multispectral TLS could enhance leaf-wood classifi­ data while carried by the operator thereby reducing understory occlu­
cation and additionally allow measurement of biochemistry and health sion (Bauwens et al., 2016) but until recently had a limited range (Cabo
to reveal the 3D distribution of canopy physiological features (Danson et al., 2018) and large beam divergence compared to stationary TLS
et al., 2014, 2018; Douglas et al., 2015; Elsherif et al., 2019b; Gaulton systems. New MLS instruments now offer an increased range up to
et al., 2013; Junttila et al., 2019). However, multispectral TLS remains 100 m (ZEB Horizon) and recent developments in MLS (Hyyppä et al.,
experimental with research instruments such as SALCA (Danson et al., 2020) are promising, but require further testing in structurally complex
2018; Schofield et al., 2016), DWEL (Douglas et al., 2015; Li et al., forests.
2016, 2018) or hyperspectral TLS (Hakala et al., 2012). The use of two Overall, the diversity of TLS systems available has produced a range
or more individual single-wavelength TLS systems operating at dif­ of novel measurement approaches, useful for characterising forest
ferent wavelengths (but with similar technical specifications) has been structure in 3D. Each system can effectively be deployed in a range of
shown to be successful for detecting differences in leaf water content or scenarios highlighted in Table 1, but, at a given price point and set of
tree health (Elsherif et al., 2019a; Junttila et al., 2019), but is compli­ specifications, certain instruments excel at specific tasks. Granted, we
cated by the need for careful radiometric calibration and beam align­ recognise many instruments can be used to collect similar 3D structure
ment. Such approaches also introduce significant logistical challenges information, especially if the limitations of each instrument are con­
in field application as multiple scans are required, especially if condi­ sidered and mitigated with appropriate acquisition strategies (e.g. re­
tions are windy or change between scans. ducing occlusion, higher density scan spacing, etc.; Wilkes et al.
Other terrestrial laser scanning instruments offer atypical acquisi­ (2017)). Reliable automation with weatherproof designs will make high
tion strategies (transecting, handheld mobile, and multi-spectral) that frequency seasonal TLS collections possible for phenological studies
enable unique perspectives of forest structure and function. Portable and change detection (Culvenor et al., 2014). Price is a major barrier for
canopy LiDAR (PCL) is an upward facing pulsed TOF that collects adopting TLS technology into forest ecology, but instruments with ac­
transects of vertical forest structure and can help capture structure- ceptable specifications are becoming available at much lower price

3
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

points. At present, commercial TLS units and onboard detection and given voxel (Pimont et al., 2018). In reality, beam divergence and the
filtering algorithms are not optimised for capturing vegetation struc­ finite number of laser pulses entering any given voxel due to angular
ture. Now, major technological advances specific to forest ecology must scanning resolution and occlusion results in voxels with varying point
address the issue of more reliably sensing “soft” surfaces (e.g. leaves) density. This will influence the choice of right voxel size (Pimont et al.,
with lower beam divergence and greater sensitivity to low reflectance 2018).
surfaces. Whereas both pulse-based and voxel-based approaches have their
own set of problems related to theoretical assumptions that cannot be
2.2. Towards algorithms for automated TLS data processing met in measured TLS data, there are some issues common to both of
them. First, to calculate LAI we need robust algorithms to distinguish
The past two decades have seen significant progress in the devel­ between leaf and woody points. We have recently seen an increase in
opment of near-automated processing pipelines for extracting different the number of leaf-wood separation methods from TLS data based on
forest structural attributes from TLS data. However, most TLS ap­ both intensity (Béland et al., 2014b) and geometric properties of the
proaches in forest ecology often still rely on some time-demanding points in 3D space (Béland et al., 2014a; Belton et al., 2013; Boni Vicari
manual steps for data analysis. The need to advance and automate al­ et al., 2019a; Krishna Moorthy et al., 2019b; Wang et al., 2018; Yun
gorithms for deriving structural features from 3D data is equally im­ et al., 2016; Zhu et al., 2018a). Methods based on geometric properties
portant as sensor advancement for the broader uptake of TLS in forest of the points are more robust than intensity-based methods as they are
monitoring. independent of the TLS instrument. However, most of these methods
are developed and tested only on temperate forests (except for Boni
2.2.1. Current approaches Vicari et al. (2019a) and Krishna Moorthy et al. (2019b)) and do not
Methods for analysing TLS data from forests can be broadly classi­ provide sufficient detail to reproduce or benchmark the results of dif­
fied into two main categories: (1) gap probability methods and (2) ferent methods. Second, both pulse-based and voxel-based approaches
geometrical modelling (Newnham et al., 2015). Pulse-based or voxel- make theoretical assumptions about the foliage distribution, which are
based gap probability methods are used to estimate plant area index not always valid in reality. Leaf angle distributions (LADs) influence the
(PAI, Fig. 1) or leaf area index (LAI) of forest stands, whereas geome­ incoming radiation regime within the canopy and are an important
trical modelling allows for explicit reconstruction of individual tree parameter in estimating vertical LAI profiles (Ross, 1981; Wilson,
structure. 1959). LAD is often ignored in ecological models due to the difficulty in
quantifying this parameter. Recent studies have succeeded in esti­
mating LAD at individual tree level from TLS data (Boni Vicari et al.,
2.2.1.1. Gap probability methods. LAI quantifies the area of leaf material
2019b; Itakura and Hosoi, 2019; Kuusk, 2020; Liu et al., 2019) enabling
per unit area in an ecosystem, and critically contributes to the
reliable estimation of LAI and vertical LAI profiles from TLS. Irrespec­
characterisation of Earth's climate (Asner et al., 2003; Calders et al.,
tive of some of the still prevalent issues in estimating PAI from TLS data
2015b; Jonckheere et al., 2004). It is important to note that TLS can
and the difficulty to measuring true PAI, recent studies have demon­
essentially only estimate PAI or WAI (wood area index) in forests
strated that TLS provides a more stable estimate of PAI when compared
(Calders et al., 2018b). Gap probability estimates are the basis for
to other ground-based sensors such digital hemispherical photography
deriving PAI and the vertically resolved plant area volume density
(DHP) as it is independent of the illumination conditions (Calders et al.,
(PAVD) based on a form of the Beer-Lambert's law (Calders et al., 2014;
2018b; Hancock et al., 2014).
Jupp et al., 2009; Pimont et al., 2018; Zhu et al., 2018b):

Pgap ( , z ) = e G ( ) PAI (z )/ cos ( )

(1) 2.2.1.2. Geometrical modelling. Geometrical modelling approaches
exploit the full 3D structure of the TLS data and generally require
z is the height above terrain and where zmax is the height of the canopy single trees to be segmented from co-registered point clouds. Multiple
and θ is the zenith angle of the laser pulse. Pgap(θ, z) is the vertically single scans can be co-registered using high reflectivity targets that act
resolved gap probability and G(θ) is the foliage orientation function, as tie-points between different scan locations (Wilkes et al., 2017).
which equals the projection of a unit area of plant constituents on a Work on reflector-less registration algorithms is promising (Kelbe et al.,
plane perpendicular to the direction θ, averaged over elements of all 2016) and a new range of commercial scanners (Leica BLK360 and
orientations (Ross, 1981). RIEGL VZi-series, see Section 2.1) provide onboard registration without
From a single scan, pulse-based methods (Calders et al., 2014; Jupp the need for targets. Currently, we recommend more testing (i.e.
et al., 2009) approximate vertically resolved gap probability, quantifying the effect of ecosystems, instrument characteristics and
Pgap ( , z ) as sampling design) before using this in an operational context. A uniform
wi (z i < z, ) point density is recommended to provide consistent point cloud quality
Pgap ( , z ) = 1 , where w = 1/ ns throughout the plot. This may require downsampling the point cloud
N( ) (2)
(e.g. using voxel grid filtering, Burt et al. (2019)) or scanning an area
is defined as the mid-point of the finite zenith angle interval used to larger than the plot (Wilkes et al., 2017).
aggregate laser pulses. The numerator in Eq. (2) gives the number of In the past decade a range of methods have been developed to ex­
laser returns that are below z and N ( ) is the total number of outgoing tract trees (Burt et al., 2019; Raumonen et al., 2015; Trochta et al.,
laser pulses for the zenith angle interval. For a specific emitted laser 2017; Yrttimaa et al., 2019a), lianas (Krishna Moorthy et al., 2019a,
pulse each return equates to a beam area interception of 1/ns, where ns 2020) or downed dead wood (Yrttimaa et al., 2019b) from plot-level
is the number of total returns for that emitted laser pulse. This approach TLS data in a (semi-)automated manner. Most tree segmentation
is implemented in the open source python library pylidar (www.pylidar. methods follow a bottom-up approach by first identifying the potential
org). stem bases and subsequently growing the identified stems into branches
Alternatively, several voxel-based methods to estimate PAI are and twigs to reconstruct full tree crowns iteratively. For example, 3D
available (Pimont et al., 2018). These include ray-tracing methods FOREST (Trochta et al., 2017) first divides the whole point cloud into
based on contact frequency approach (Béland et al., 2011) and methods horizontal slices and further divides each of these slices into clusters
based on Beer-Lambert's law (Béland et al., 2014b; Grau et al., 2017; based on user-defined parameters (e.g. cluster size and maximum dis­
Hosoi and Omasa, 2006). Voxelising TLS data is not trivial especially tance between two points to belong in the same cluster). Each of the
when we deviate from the theoretical assumptions that TLS instruments clusters are treated as a potential tree and are merged vertically with
emit an infinite number of infinitely small laser pulses through any the closest clusters from other slices based on the angle and distance

4
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

Fig. 2. Conversion of a segmented point cloud (Burt et al., 2019) from terrestrial laser scanning to a virtual forest for radiative transfer modelling, example of 1 ha
Wytham Woods (Calders et al., 2018a). (a) & (c) give a view from above of the 3D point cloud and 3D model respectively. (b) & (d) give a side view of a single tree
point cloud and its corresponding model (branches + leaves) respectively.

between the centroids of the clusters. Treeseg (Burt et al., 2019) follows (Goodfellow et al., 2016).
a slightly different approach by first detecting the stem points closer to While object detection, classification and segmentation in 2D
the ground instead of treating every cluster as a possible tree. The tree images have moved away from classical ML to deep learning (Toshev
is then further extracted using generic point cloud processing techni­ and Szegedy, 2014), segmentation of 3D data is still predominantly
ques including Euclidean clustering, principal component analysis, re­ based on classical CV and ML algorithms. The first set of deep learning
gion-based segmentation, shape fitting and connectivity testing. networks for 3D data projected the 3D point clouds into 2D images from
Fig. 2(a-b) shows a segmented point cloud using treeseg. These methods multiple viewpoints and used 2D convolutional neural networks
usually require manual assistance and quality control to correct omis­ (Rehush et al., 2018; Su et al., 2015). Deep learning for 3D point clouds
sion or commission errors in the segmented point clouds. Generally, has made considerable progress and evolved from converting point
more manual intervention is required in complex ecosystems, such as clouds to voxels (Xi et al., 2018; Zhou and Tuzel, 2018) or octrees
tropical rainforest, where multiple crowns can interact with each other. (Riegler et al., 2017) to working directly on 3D point clouds (Qi et al.,
Once the trees are extracted, the tree point cloud can be modelled 2017a, 2017b). Furthermore, a recent deep learning based network
using quantitative structure modelling (QSM) algorithms. However, called BranchNet has been trained to specifically extract structural in­
leaf-wood separation might be required first for leaf-on point cloud formation from branch-like structures (Halupka et al., 2019).
data. Current state-of-the-art leaf-wood separation algorithms are Deep learning techniques can potentially automate the processing
mostly based on machine learning (ML) and computer vision (CV) ap­ pipeline for extracting various features from 3D data of forests. This
proaches (Béland et al., 2014a; Belton et al., 2013; Boni Vicari et al., includes, but is not limited to, segmenting individual tree stems and
2019a; Krishna Moorthy et al., 2019b; Wang et al., 2018, 2020a; Yun branches (Xi et al., 2018) to extracting detailed branch structural in­
et al., 2016; Zhu et al., 2018a). Classical ML algorithms such as random formation (Halupka et al., 2019). However, challenges still need to be
forests and Gaussian mixture models (GMM) were trained on manually overcome. The main success of deep learning in 2D image segmentation
labelled point clouds from specific forest types with features based on is the availability of a large number of images to train and benchmark
eigenvectors and eigenvalues that describe the local geometric prop­ these networks. Reference datasets of 3D tree architecture that would
erties of the points in 3D space. Given that the training data points for a be useful for training and benchmarking the algorithms are currently
model come from a handful of forest types, further testing of these al­ lacking. This is not surprising considering the difficulty in creating re­
gorithms is advised. Unsupervised approaches are preferred over su­ ference datasets. Open access to these already available reference da­
pervised ML approaches considering the difficulty in acquiring manu­ tasets with well-described metadata and uncertainties would facilitate
ally labelled data points (Wang et al., 2020a). deep learning approaches for TLS. With the increasing amount of TLS
Tree structural metrics related to branching architecture can be data being collected across the world, deep learning based algorithms
derived through skeletonising methods that derive a graph with geo­ have the potential to revolutionise the field (Arel et al., 2010) and could
metric information of the vertices and edges from the point cloud fuel the automation of some of the existing manually intensive tasks of
(Bucksch and Lindenbergh, 2008). Based on QSM algorithms such as extracting features from the 3D data.
TreeQSM (Calders et al., 2015a; Raumonen et al., 2013) or simpletree
(Hackenberg et al., 2015) that both fit cylinders, the tree volume as well
3. Forest measurement and management
as 3D structural metrics and their topology can be estimated. The
quality of the QSM depends on the quality of the point cloud data and
Key observation variables in forest monitoring and management are
quantifying uncertainty of QSMs remains challenging. Whereas the
DBH, tree height, basal area per hectare, stand growing stock volume,
QSM cylinder fitting approach might work for most trees (Akerblom
and aboveground biomass (AGB). In order to generate these attributes,
et al., 2015), it might fail for buttressed trees in tropical forests (Disney
individual trees need to first be identified providing an attribute of tree
et al., 2018) and mesh-based models are advised (Liski et al., 2014;
count. Henning and Radtke (2006), Maas et al. (2008) and Liang et al.
Morel et al., 2018). Open source initiatives such as computree (Othmani
(2012), to name a few, presented methods for identifying tree stems
et al., 2013) and 3D FOREST (Trochta et al., 2017) have integrated the
from TLS point clouds whereas Donager et al. (2018) compared point
point cloud processing workflow in a GUI.
densities and Liang et al. (2018) compared a variety of methods in tree
detection. Both Donager et al. (2018) and Liang et al. (2018) concluded
2.2.2. The potential of deep learning for TLS data processing that increasing forest density increased the challenge of correctly de­
Deep learning differs from classic ML algorithms in how features are tecting all trees within the area of interest. Reliable tree count estimates
extracted from data. In classical ML, features are handcrafted by hu­ from TLS have implications for the reliability of data fusion and up­
mans and then fed into classification algorithms, whereas in deep scaling applications (see Section 4).
learning, the algorithm learns the features by itself from the data Circle or cylinder fitting have most commonly been used for

5
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

deriving DBH estimates from TLS (Liang et al., 2018). Nevertheless, expected to be less transferable, due to the differences in site-specific
both circle and cylinder fitting assume circularity of tree stems, which is tree allocation patterns across environmental gradients. We recommend
a rather optimistic assumption (Saarinen et al., 2014; Stovall et al., that it is essential to carefully evaluate (potential) empirical models that
2017) and could further be enhanced as TLS data enables modelling of use TLS data to estimate AGB as well as tree allocation patterns in
more complex primitives (Akerblom et al., 2015) or convex hull different ecoregions to better understand forest dynamics and especially
(Stovall et al., 2017). TLS estimates of tree height from co-registered suitability of tree allometric models across forest types.
point clouds (Wilkes et al., 2017) have been obtained as a difference Crown structure is difficult to measure automatically and objec­
between the highest and lowest points of tree point clouds (Calders tively with traditional forest measurement devices but can be extracted
et al., 2015a; Saarinen et al., 2017) or as the value of 99.9th percentile from TLS data (Kankare et al., 2013; Lau et al., 2019a; Srinivasan et al.,
of height (Stovall et al., 2017). Several studies have reported under­ 2015). For instance, Seidel et al. (2011) and Metz et al. (2013) have
estimates for TLS-based tree height (reviewed by Liang et al., 2016, presented a variety of attributes derived from TLS point clouds char­
Liang et al., 2018), although accurate height measurements have also acterising crown size and shape (e.g., crown height, projection area,
been reported when compared with destructively felled trees (Calders volume, asymmetry) using a convex hull of points in a plane at various
et al., 2015a; Stovall et al., 2017). These different observations in the heights as well as a 3D convex hull for calculating crown surface area.
accuracy of estimating tree height can be attributed to differences in Literature on the accuracy of TLS-based crown attributes is not ex­
sensors and fieldwork setup (see Section 2.1). A combination of TLS tensive (Fleck et al., 2011; Seidel et al., 2015). Fleck et al. (2011)
with 3D observations above canopy can be used to enhance tree height compared the reliability of crown area estimates from TLS with the area
estimates (Schneider et al., 2019; Yrttimaa et al., 2020). However, this of eight-point crown projections measured in the field and obtained R2
approach warrants more research in the future, especially within more of 0.96 and RMSE of 6.5 m2 whereas Seidel et al. (2015) reported
complex forest ecosystems. Additionally, scan design (i.e. number and correlations from 0.5 to 0.7 between crown attributes measured with
location of scans) as well as scanner technology (i.e. time-of-flight vs traditional means (e.g. clinometer, densiometer, measuring tape) and
phase-shift, single return vs multiple returns) should more thoroughly derived from TLS. However, in dense forests where crowns interweave,
be investigated in order to better understand effects of each component reliably obtaining crown attributes from TLS point clouds is similarly
on tree height accuracy. challenging to field measurements. This requires further methodolo­
Both stem volume and AGB are traditionally indirectly estimated gical development, which can also contribute to automatically and re­
using allometric models with field measurements such as DBH and tree liably segmenting individual trees (see Section 2.2.1.2).
height as predictors (Chave et al., 2005; Henry et al., 2011; Zianis et al., Taper curves provide diameters along a stem and it has been utilised
2005). Current TLS approaches reduce allometry-related uncertainty for obtaining stem volume, especially in Scandinavia. Measuring dia­
through direct estimates of woody volume from point clouds (Gonzalez meters from the upper part of a stem (i.e. within crown), especially for
de Tanago et al., 2018; Kankare et al., 2013; Yu et al., 2013), which can conifers, can be challenging from TLS point clouds. A spline function
be converted to AGB using wood specific gravity (WSG) information. (i.e. a proxy for taper curve) can be used in completing the diameter
For carbon stock assessments, proper error accounting is crucial, measurements from the occluded part of the stem (Saarinen et al., 2017,
especially for large trees for which current allometric models are most 2019). This approach can also be used for buttressed trees, with the
uncertain (Case and Hall, 2008; Chave et al., 2014; Réjou-Méchain overall shape of the stem being estimated from the upper well-scanned
et al., 2019). Conversion of volume into AGB through WSG adds un­ part downwards with the use of a taper curve (Bauwens, S. et al., un­
certainties to AGB estimates, caused by the high spatial, intra-specific published).
and intra-individual variability of WSG, and the occurrence of hollow These new developments in forest measurement and new structural
stems, which cannot be detected by TLS. Often it is impossible to metrics from TLS are relevant to forest management. For example, TLS
sample each scanned tree for WSG, hence approximative values are provides new information on how structural crown properties vary in
sourced from species-specific or plot WSG averages. This requires the mixed and pure stands (Barbeito et al., 2017; Bayer et al., 2013; Hajek
availability of sufficient WSG data coupled to correct species identifi­ et al., 2015; Kunz et al., 2019) as well as how canopy gaps impact
cations. Åkerblom et al. (2017) showed that tree species recognition crown shape (Hess et al., 2018; Seidel et al., 2016). Detailed informa­
with QSMs is feasible in low-diversity boreal forests. For more species tion about the taper curve (Pitkänen et al., 2019; Saarinen et al., 2017)
diverse forests, tree species classification performance greatly depends provides information on log geometry and wood quality (Pyörälä et al.,
on the targeted application and exhibits a trade-off between sensitivity 2019a). Future improvements of forestry volume tables will be possible
and specificity (Terryn et al., 2020). Ideally, a whole-tree or volume- by repeated TLS measurements to gain new insights into growth
weighted WSG is used as a conversion factor. Discrepancies between a (Mengesha et al., 2015; Sheppard et al., 2016), changes in stem taper
database WSG value, or a partial WSG measurement (e.g. increment (Luoma et al., 2019) and biomass (Kaasalainen et al., 2014; Srinivasan
coring at breast height) and whole-tree WSG can contribute to bias in et al., 2014). Repeated TLS data acquisition requires careful planning
TLS-derived AGB estimates (Sagang et al., 2018; Wassenberg et al., for ensuring point clouds with comparable quality. Co-registering of
2015). Efforts to derive whole-tree WSG from a partial WSG measure­ multiple point clouds from different time points should result in a si­
ment have been developed to some extent (Bastin et al., 2015; Momo milar level of quality. Furthermore, scan design and geometry should
Takoudjou et al., 2020; Wassenberg et al., 2015). Other solutions to provide comparable information on the same trees, crowns, and bran­
mitigate WSG related uncertainties, such as novel WSG sampling ches for reliably assessing possible changes and their magnitude. An
methods, are discussed in (Réjou-Méchain et al., 2019). example of intensive temporal resolution of a TLS time series for which
The increased accuracy of direct volume and derived AGB estimates these challenges were minimized was presented by Puttonen et al.
with TLS data is poised to improve the quality of carbon stock assess­ (2019) who demonstrated circadian movements of tree branches during
ments through permanent sample plots, supersites or national forest a 14.5 h measurement period.
inventory (Liang et al., 2016). Alternatively, TLS can be used to include Silvicultural practices affect growing conditions (i.e. light, tem­
information about crown structure into allometric models to better perature, water, nutrients) of individual trees and thus, forest structure
distinguish the heteroscedasticity of tree size-to-mass allometry and tree growth (Eriksson, 2006; Juodvalkis et al., 2005; Mäkinen and
(Goodman et al., 2014; Kankare et al., 2013; Lau et al., 2019a; Ploton Isomäki, 2004; Nilsson, 2010; Río et al., 2017; Valinger et al., 2019).
et al., 2016). Whereas existing pantropical tree allometric models are Since TLS can provide a variety of structural attributes of trees as well
transferable across tropical forest types (Chave et al., 2014), more as their relationship (e.g. spatial distribution or crown competition),
complex models, including additional geometric plant features from which have traditionally been demanding or impossible to measure,
TLS, may show a better performance locally. However, they are TLS can expand our understanding about the effects of silviculture. TLS

6
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

has already been used in investigating effects of silviculture practices on measurements of branch architecture can test not only the predictions,
structural diversity through space filling (Juchheim et al., 2017a), but also the assumptions, of how tree structure and metabolism scale
competition and tree structure (i.e. stem and crown attributes) (Georgi with size (Metabolic Scaling Theory, MST, West et al. (1997) and
et al., 2018; Juchheim et al., 2017b). Additionally, TLS has shown its Savage et al. (2010)). Testing these assumptions holds large implica­
potential in providing information on how species composition affects tions for the use and potential revision of MST predictions related to the
tree architecture (Barbeito et al., 2017; Bayer et al., 2013; Juchheim scaling of plant growth (Enquist et al., 2007) and to linking forest
et al., 2020; Krůček et al., 2019; Metz et al., 2013). However, the ma­ structure and productivity (Enquist et al., 2009; Fyllas et al., 2017; West
jority of studies on the effects of silviculture have been concentrating on et al., 2009). For example, this theory assumes that branch radii and
deciduous trees, especially European beech (Fagus sylvatica L.) in cen­ branch lengths follow a scaling relationship, but testing this assumption
tral Europe. Therefore, we encourage the use of TLS also for conifers accurately would require felling trees and measuring the size of branch
and in boreal forests where forest management and silviculture have segments by hand, a prohibitively time consuming task for large trees
long traditions to understand the potential of TLS in various forest (Bentley et al., 2013). As such, Lau et al., 2018 proposed a method to
environments (Saarinen et al., 2020). use TLS to quantify fine scale branch architecture. A proof of concept
In addition to forest management, TLS can be used in mapping and from nine sampled trees in Guyana was deemed successful to test
measuring downed dead wood (Yrttimaa et al., 2019b) - an important branch scaling predictions of MST (Lau et al., 2019b).
attribute for biodiversity-, classifying defoliation (Huo et al., 2019), Furthermore, TLS can measure the quantitative architectural metric
characterising tree health (Junttila et al., 2019), and natural sway path fraction. This quantity is the ratio of the mean path length (from
frequencies (Jackson et al., 2019) related to tree architecture and tree- tree base to branch tip) to maximum path length (Smith et al., 2014),
wind dynamics. Understory vegetation, obtained by TLS, can be related and it relates to tree hydraulic efficiency. Relatively little research has
to regeneration or forest structural complexity (Willim et al., 2019). been devoted to the reconstruction of fine scale branching architecture,
Microstructure (i.e. vegetation canopy and topography at cm to m and the estimation of the metric path fraction thus remains challenging
scales, Maguire et al. (2019)) has been shown to affect photosynthetic for trees (Lau et al., 2018). Occlusion and wind effects while scanning
functioning in forest-tundra ecotone with TLS, and a relationship be­ have been well established to be problematic (Seidel et al., 2012; Vaaja
tween rainfall interception and LAI was demonstrated by (Yang et al., et al., 2016; Wilkes et al., 2017), but are even more so when fine
2019). Finally, TLS has enabled assessing forest surface fuel loads (Chen branches need to be estimated. In addition, scanning to the level of
et al., 2017; Wallace et al., 2016) and it is expected to enhance forest precision needed to estimate small branches requires instruments with
fuel hazard assessments as well as planning ecological and risk miti­ small beam divergence, extremely high-resolution scans, and a small
gation strategies. scan position grid (~10 m) (Wilkes et al., 2017), which are time con­
suming to acquire. Recent developments in measuring small fine-scale
4. New opportunities with terrestrial laser scanning branches include the use of optical photogrammetry to potentially re­
fine TLS size estimates for small fine-scale branches towards the tips of
Beyond the applications of TLS in the fields of forest measurement trees (Wilkes et al., 2019); and the development of modified QSMs that
and management there is a wide range of new application areas where combine leaf and wood processing (Boni Vicari et al., 2019a; Krishna
tree architectural information, at a range of scales, may provide the key Moorthy et al., 2019b; Wang et al., 2020a) or allow data co-registration
to new scientific insights (Disney, 2019). For example at the finest and volume enclosure to be considered as part of the same process
spatial scales, TLS can provide direct measurements of tree canopy (Wang, 2020).
components like branch size, position and orientation, all of which are
impractical to measure manually. These measurements can provide 4.2. Virtual tree models for radiative transfer models
fundamental new insights into the scaling of branches. Furthermore,
this feeds into reconstruction of virtual trees that can be scaled up to Radiative Transfer Models (RTMs) are an integral tool within Earth
represent complete forest stands (Calders et al., 2018a). These virtual Observation (EO) of forests, as they enhance our ability to monitor and
forest models can in turn be used to improve the accuracy of the re­ understand the linkages between light emitted or reflected (i.e. an EO
trieval of forest biophysical properties from radiative transfer models observed signal), forest structure and biochemistry. Representing forest
(RTMs) when coupled with Earth Observation (EO) data. New appli­ structure in RTMs can range from 1D horizontally homogeneous layers
cations are rapidly developing to characterise and quantify the bio­ to fully geometrically explicit 3D models, where the latter can facilitate
physical properties of urban trees (Baines et al., 2020). Recent research the calibration and validation of EO data, better uncertainty quantifi­
has illustrated the magnitude of the contribution of such trees to carbon cation and algorithm development.
storage and other ecosystem services (Wilkes et al., 2018). A final ex­ Previously, 3D virtual forests were reconstructed using softwares
ample of new insights arises through the opportunity to obtain accurate such as OnyxTree (www.onyxtree.com), xfrog (Lintermann and Deussen,
repeated measurements with TLS to detect landscape-scale ecosystem 1999) or arboro (Weber and Penn, 1995), which requires para­
changes (Singh et al., 2020). Long-range TLS measurements may be meterisation from field inventory data, airborne LiDAR, or parametric
limited in terms of spatial resolution at kilometer scales but recent re­ modelling of plant growth and topology. However, numerous gaps in
search has highlighted the power of long-range measurements for information based on traditional inventory data exist, which can (po­
ecosystem change detection. These four examples are explored in more tentially) be addressed by TLS, including: shoot/leaf shape and di­
detail next. This review focuses on forest ecosystems, but it is worth mensions (curl, size) (Åkerblom et al., 2018; Zhu et al., 2018b), shoot/
mentioning that TLS data is increasingly used in horticultural tree crops leaf orientation (Boni Vicari et al., 2019b), foliage distribution in
for assessing tree structure (Decuyper et al., 2018; Fernández-Sarría crowns (Martin-Ducup et al., 2018), crown shapes (Côté et al., 2009)
et al., 2019; Moorthy et al., 2011) and advancing the development of and the wooden skeleton including branching angles and density
monitoring with remote sensing platforms (Wu et al., 2020a, 2020b). (Calders et al., 2018a; Raumonen et al., 2013). Fig. 2(c-d) shows an
example of a fully explicit structural model derived from TLS where the
4.1. Metabolic scaling theory woody components are modelled by QSMs and leaves are added with
the FaNNI algorithm (Åkerblom et al., 2018) through LAI estimates
Measurements resulting from TLS can provide precise and detailed from TLS.
information on tree structure that can help answer fundamental ques­ Despite these advancements, a number of challenges still remain,
tions about tree size and shape, allometric scaling, metabolic function particularly in providing a fully explicit representation of individual
and plasticity of form (Disney, 2019). Importantly, detailed TLS leaves/shoots and/or tree crowns (for example recent work by Wang

7
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

et al. (2020b) demonstrated promising results on the use of TLS to imagery or LiDAR sensors on (driverless) cars. Further, citizen science
extract photon recollision probability at the crown level, allowing de­ projects (e.g. Treezilla) could also be used as additional training data or
lineation of tree crown structures), understory structure and woody validation for remote sensing derived urban forest structure or deep
content (live and dead) (Lau et al. (2018)). Along with these challenges, learning approaches.
further considerations need to be taken into account when using TLS
measurements for RT modelling. The collection of ground-based data 4.4. Long-range scanning
concurrently with TLS is necessary for RT modelling, including the
scattering directionality of the forest elements (leaves/shoots, wood, Most progress in ecological TLS research has taken place in forested
understory and background bidirectional reflectance distribution ecosystems, where the laser ranging distance of an instrument is seldom
function) and illumination, as well as in situ measurements of EO data a limiting factor for long-range TOF instruments. Line-of-sight is typi­
products (e.g. LAI, Fraction of Absorbed Photosynthetically Active Ra­ cally limited to < 100 m in temperate, boreal and tropical forests
diation or biochemical constituents) to validate reconstructed virtual (Calders et al., 2014), and avoiding excessive occlusion requires multi-
forests and simulated EO measurements of forest structure and function scan setups with 10 or 20 m grid spacing (Wilkes et al., 2017). In these
(Calders et al., 2018a; Cifuentes et al., 2018; Widlowski et al., 2014). forests there is little benefit to be gained by investing in laser systems
Site selection is also important, specifically if TLS measurements are that can range accurately over distances longer than 1 km. As such, the
used to characterise calibration and validation networks, such as the spatial extent of TLS data acquisitions is typically restricted to < 1 ha in
Terrestrial Ecosystem Research Network (TERN) (Karan et al., 2016), scale (Beland et al., 2019), limiting their suitability for monitoring
the National Ecological Observatory Network (NEON) (Kao et al., hillslope and landscape scale dynamics. Outside of forestry research
2012), and the Integrated Carbon Observation System (ICOS) (Gielen however, much progress has been made in developing long-ranging
et al., 2018) to name but a few. Both larger sites, and consistent mea­ TOF TLS instruments (> 2000 m) for applications in glaciology, geo­
surements over time are needed, in order to represent the spatial het­ morphology, archeology, and mining engineering (Fischer et al., 2016;
erogeneity (Morsdorf et al., 2020) and temporal evolution (Calders Gabbud et al., 2015; Lercari, 2016). Long-range scanning offers po­
et al., 2018a) of forests as seen by a remote sensing sensor. Finally, data tential to complement the TLS methodologies that have been developed
fusion of TLS with hyper- or multi-spectral remote sensing data can in forests so far, but their suitability for ecological studies will vary as a
improve sampling of the distribution of canopy spectral properties to function of the habitat type and landscape terrain features.
better parameterise the spectral information input into RTMs, as well as The utility of long-range scanning to inform ecological questions,
providing additional data sources for validation of the RTMs, for ex­ such as vegetation response to disturbances, has only recently been
ample Schneider et al. (2014) reconstructed a virtual forest based on explored (Singh et al., 2018), and provides an avenue for obtaining
ALS- and TLS-derived voxel grids of PAI, and showed good agreement high-resolution 3D data at hillslope scales (Singh et al., 2020). Long-
between simulated and measured hyperspectral data from the APEX range TLS is suited to open landscapes with uninterrupted line-of-sight
airborne imaging spectrometer (Schaepman et al., 2015). Addressing for hundreds of meters (Fig. 4). These criteria can be met in many sa­
such challenges can improve realistic radiative transfer modelling of vanna and shrub-land ecosystems around the globe, especially if the
forests and help provide a better understanding of the interaction be­ TLS instrument can be elevated to positions above the tree canopy and
tween forest structure and EO derived parameters. the topography is relatively flat. A key consideration of long-distance
scanning is the beam divergence of the laser. For example, the RIEGL
4.3. Trees outside of forests VZ-2000 has a beam divergence of 0.35 mrad, which translates to a
footprint spot size of 0.04 m at 100 m distance, and 0.7 m at 1000 m.
TLS has been applied to quantify the structure of trees outside of Explicitly testing how beam divergence and incidence angle affect the
forests as well, for example in commercial orchards (Murray et al., characterisation of vegetation canopies are key areas of ongoing re­
2020; Wu et al., 2020a) and as discussed below in urban forests. The search, which will lead to better quantification of how error propagates
ecosystem services offered by urban forests are now recognized for with distance from the scanning position. These advances will allow for
making our ever expanding cities more habitable for city dwellers in the broader scale 3D mapping that can capture the structure and dynamics
face of changing climate. The structure of urban trees and forests can be of heterogeneous systems, which are difficult to represent through
highly variable due to context, management and planting strategies traditional plot- or transect-based scanning approaches.
which often deviate from natural forests. This can lead to structural
outliers (e.g. very tall or open grown) and may give insight into the 5. Towards global ecosystem understanding
limits of growth for particular species (Disney, 2019). TLS allows for
capturing variability in tree morphology when compared to traditional Local measurements are useful for key challenges such as carbon
measurement techniques; this includes capturing environmental con­ balance or long-term forest monitoring only if they can be scaled up to
text (e.g. tree spacing, buildings, etc.). TLS datasets can therefore be the ecosystem, landscape, and regional scale. The extensive TLS data
used to derive new allometric models specifically for quantifying the collection efforts across diverse forest ecosystems are increasingly being
volume of urban trees, for example, to estimate AGB (Lefsky and coupled and augmented with airborne and spaceborne LiDAR char­
McHale, 2008; Wilkes et al., 2018). Further, TLS is an excellent way to acterisation of vegetation structure. This facilitates appropriate scaling
digitise, study and monitor exceptional or unusual trees which are often from individual trees to local plot measurements to regional estimates.
found in cities (Fig. 3). TLS data is expected to play a crucial role in helping revise and extend
The use of TLS in urban forestry is not yet operational, and still ecological scaling theories related to tree form and function to help
requires method development especially for trees outside of forests. determine the ecological and evolutionary drivers (Magney et al.,
However, TLS datasets have been used to train predictive models of 2014). This could then further relate branch architecture traits to leaf
urban forest structure. When combined with airborne and satellite re­ and wood properties at the whole tree level (Verbeeck et al., 2019).
mote sensing data (Baines et al., 2020; Tanhuanpää et al., 2019), this However, when we aim for upscaling to stand and landscape level,
can identify patterns in urban forest structure and allow monitoring of there is a practical limit to the amount of resources that can be allocated
highly dynamic forests through time. A barrier to operational adoption to the collection of TLS data over larger areas (> 1 ha). We should
of TLS for urban forest inventory could be cost and the time taken to question if the high point density and level of detail acquired with TLS
acquire data. However, cities are some of the most surveilled areas on is always required for studies at larger spatial scales. In this context, it is
earth and offer opportunities to supplement TLS with new 3D mea­ useful to consider the potential of other laser scanning platforms
surements, e.g. Structure-from-Motion (SfM) from airborne aerial (spaceborne, airborne, Unmanned Aerial Vehicles - UAV), and evaluate

8
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

Fig. 3. TLS data of the Hardy tree, situated in the grounds of St Pancras Old church in London (51.5350° N, 0.1302° W). This Ash tree (Fraxinus excelsior L.) is notable
for the many gravestones sitting leaf-like in around the base of the tree, that were supposedly placed there by the author Thomas Hardy in the 1860s during his time
working as a clerk on the rapid expansion of the railway system. The tree is senescing and is being actively managed by Camden Council (compare summer 2017 and
winter 2019). The tree has been scanned three times with a RIEGL VZ-400 to monitor how tree structure changes over time. An interactive 3D model can be viewed at
https://skfb.ly/6GVBK.

the added value of fusing data from these platforms with TLS data. Fig. 5 demonstrates different point densities for ALS, UAV-LS and TLS
Unmanned Aerial Vehicles equipped with laser scanners (UAV-LS) along a 150 m transect in a tropical savanna.
have been explored as a possible solution to speed up the scanning UAV-LS has been successfully used for a number of forestry related
process, in order to cover larger areas and still allow analyses com­ applications. These applications include: tree height estimation and
parable to those from TLS (Brede et al., 2017; Liang et al., 2019; localisation (Wallace et al., 2014b), tree detection and segmentation
Wallace et al., 2014b). There are currently multiple commercial UAV (Balsi et al., 2018; Wallace et al., 2014a), DBH estimation (Brede et al.,
systems available, with a large variation in data quality. Recent UAV-LS 2017; Wieser et al., 2017), Canopy Height Model (CHM) generation,
systems have produced point clouds with densities of around 50 LAI estimation, AGB estimation via allometric equations based on tree
(Wallace et al., 2012), 1500 (Gottfried et al., 2015; Jaakkola et al., height and crown area (Guo et al., 2017), and tree parameter estimation
2010) and 4000 points per m2 (Brede et al., 2017). Compared to tra­ from tree reconstruction algorithms (Brede et al., 2019). We emphasise
ditional Airborne Laser Scanning (ALS), UAV-LS demonstrates sig­ again that QSMs only estimate volume and conversion to AGB through
nificantly higher point density at lower cost and with higher flexibility, wood specific gravity will introduce additional uncertainties (see
but with significantly smaller spatial coverage. The choice for ALS or Section 3). A comparison between the TLS and UAV-LS systems from
UAV-LS mainly depends on the size of the study area and application. RIEGL show that for more open forest types (temperate beech forest or

Fig. 4. Long-range scanning. Panel (a) shows the

RIEGL VZ-2000 setup on a topographic vantage
point in southern Kruger National Park, South
Africa. Panel (b) shows the point cloud difference for
a cross-section (2015–2016). Blue points are un­
changed, red points are not present anymore (e.g.
red tree on the left has been toppled, mostly likely by
an elephant). Yellow/green also indicates vegetation
loss, but at a small magnitude of less than 0.2 m (e.g.
defoliation). No growth was detected in this example
due to extreme drought conditions. (For interpreta­
tion of the references to colour in this figure legend,
the reader is referred to the web version of this ar­
ticle.)

9
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

Fig. 5. LiDAR data at the tropical savanna Litchfield TERN supersite in Australia. The LiDAR cross section is 150 m long and 10 m wide constructed from three
sections of 50 m × 10 m (from left to right: ALS, UAV-LS, TLS). All point cloud data is downsampled to 0.02 m voxels to give a fair comparison of point density. ALS
data was collected through TERN in June 2013, TLS data collected in August 2018 and UAV-LS data collected in September 2018. For scale, we show the footprint
size of a NASA Global Ecosystem Dynamics Investigation (GEDI) laser footprint.

savannah woodlands), the UAV-LS data of the RIEGL RiCOPTER the findings of Blair and Hofton (1999), recent work has developed
equipped with a VUX-SYS has a lower point density and distribution methods for simulation of GEDI waveforms from ALS and UAV-LS point
than the RIEGL VZ-400 TLS data. However, the two datasets are geo­ clouds and collocation of these simulations with on-orbit measurements
metrically still very comparable and QSM algorithms can successfully from GEDI (Hancock et al., 2019). These methods may be extended for
be applied on the UAV-LS data (Brede et al., 2017, 2019). Other UAV-LS simulation of waveforms from TLS point clouds (Hancock et al., 2017),
systems on the market yield a lower point density and data quality, enabling simulated spaceborne measurements to be directly linked with
which may limit possibilities to reconstruct the tree structure. the forest structure measurements described in Section 3.
Schneider et al. (2019) found that generally 71% of a canopy up to TLS and UAV-LS present unique opportunities to improve the ac­
25 m above ground may be occluded in a temperate forest when ob­ curacy of forthcoming AGB maps from these spaceborne missions and
served with UAV-LS. In more dense forest types (e.g. tropical rainforest also provide more insight into their uncertainties. Disney et al. (2019)
or coniferous forest) the above-canopy viewpoint of the UAV-LS limits outlined the following key areas where TLS would contribute:
capturing the full tree structure. To overcome this, critical future re­
search will include developing methods to: (a) upscale fine-scale • Improvements of the existing allometric models through increased
structure over larger areas by fusion of TLS and UAV-LS to benefit sample sizes and reduction in allometric bias through incorporating
studies where forest dynamics are high and occur at a limited spatial more near-direct measurements of large tree volume and stem dia­
scale (< 100 ha); and then (b) further extrapolating this to regional meter (Burt et al., 2020; Stovall et al., 2018; Vorster et al., 2020).
scales (> 100 ha) by fusion of UAV-LS and ALS data (Boucher, 2019; This is likely to reduce the uncertainty of estimates compared with
Pyörälä et al., 2019b). Fusion of LiDAR data from terrestrial with dif­ allometric methods that underpin all current spaceborne AGB esti­
ferent platforms and the use of fused point clouds has only been mar­ mates (Stovall and Shugart, 2018).
ginally explored in forest ecology (Paris et al., 2015; Shenkin et al., • Development and testing of EO retrieval using 3D RTMs (Calders
2019; Wilkes et al., 2018). Co-registration of multi-source point clouds et al., 2018a) (see also Section 4.2). A key problem of validating EO-
will be critical, but can be achieved if enough in common objects are derived products is the difficulty of making direct measurements of
present to act as tie-points (Calders et al., 2014). Particularly in dense the desired biophysical properties (Disney, 2016). New TLS-derived
forests, fusion of TLS with above-canopy LiDAR (e.g. UAV-LS, Schneider structure for RTM allows for integration of much more realistic and
et al. (2019)) can significantly reduce occlusion. detailed 3D surface structure into retrieval and testing processing
We are presently in a new era of spaceborne active remote sensing, chains (Calders et al., 2018a).
with three missions having accuracy requirements linked to their AGB • Quantifying uncertainty in retrieved AGB estimates through pre-
data products. These missions include two Synthetic Aperture Radar launch modelling and calibration, and post-launch validation at the
(SAR) instruments – ESA BIOMASS (https://earth.esa.int/eogateway/ plot scale (Réjou-Méchain et al., 2019).
missions/biomass) at P-band (Quegan et al., 2019; Toan et al., 2011) • Providing a link between measurements made at the tree and plot
and NASA/ISRO NISAR (https://nisar.jpl.nasa.gov/) at L-band (Rosen scale, from forestry, UAV-LS, ALS, and spaceborne platforms
et al., 2015) – and the NASA Global Ecosystem Dynamics Investigation (Kellner et al., 2019).
(GEDI, https://gedi.umd.edu/) LiDAR on the International Space Sta­
tion (Dubayah et al., 2020). The NASA GEDI LiDAR mission was The Committee on Earth Observation Satellites (CEOS) Land
launched on December 5th 2018 and started scientific data collection Product Validation (LPV) focus area for AGB has developed a protocol
on April 19th 2019 (Dubayah et al., 2020). GEDI will collect global for best practices in the validation and comparison of AGB map pro­
scale measurements of vertical canopy structure for two years using ducts (Duncanson et al., 2019). This protocol has outlined key un­
eight ground tracks (600 m spacing across track) composed of ~25 m certainties and knowledge gaps, and synthesized recommendations to
laser footprints (60 m spacing along track). Each of these spaceborne advance TLS and UAV-LS from a research technology to one that is used
missions have calibration and validation programs and TLS and UAV-LS more routinely in the calibration and validation programs of space­
contribute to these efforts (Duncanson et al., 2019). Fig. 6 shows two borne biomass missions and the establishment of biomass supersites
examples of GEDI waveforms and derived elevation and height metrics (Chave et al., 2019). Protocols for data product quality assessment and
from the TERN Litchfield SuperSite, where the footprint size is larger quality control, metadata and attribution are also required to see TLS
than individual trees. An illustration of the scale of these individual data products reach a similar level of maturity as comparable protocols
GEDI footprint measurements relative to the detailed 3D information in for forest mensuration and ALS in large area forest plot monitoring
the ALS, UAV-LS and TLS point clouds is shown in Fig. 5. Building on networks. Existing networks (e.g. TERN) have made substantial

10
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

Fig. 6. Two GEDI waveform examples at the tropical savanna Litchfield TERN supersite in Australia. The plots show the vertical profile of digitizer counts for a single
shot, which has a footprint diameter of nominally 25 m (see Fig. 5 for a spatial reference). The elevation of the lowest mode (“ground”) and highest reflecting return
(“canopy”) are derived from the waveform and available in the GEDI Level 2A product (Dubayah et al., 2020). The green line indicates the cumulative digitizer
counts between the elevation of the lowest and highest return. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

progress towards standardization of acquisition protocols and are col­ complexity.

lecting new plot scale measurements contemporaneous with GEDI. Finding ways to reconstruct 3D structures that better incorporate
probability in both process and outcome, should tell us significantly
more about the organisms and systems we are trying to measure.
6. Conclusions and outlook
Ongoing work in graphics on incorporating semantic info to assess and
infill occlusion for image reconstruction (Li et al., 2020) can potentially
TLS, along with other developments in 3D imaging, has already
be applied to point clouds (Miao et al., 2020; Shu et al., 2019). A key
provided a step change in our ability to measure tree structure in de­
part of this needs to be validation and benchmarking. A diversity of
tailed 3D, from the individual to the canopy-scale. As the interest in,
approaches to deriving structural information is vital, but this means
and use of TLS for forest ecology has increased, a number of new
that we also need objective ways to decide which approaches work best,
challenges are arising in order to make the best use of existing and new
and why. This is an area where the sheer difficulty of collecting detailed
data and tools. TLS opens a realm of untapped research questions and
and direct 3D structural measurements, makes validation of indirect
applications that call for the most detailed and accurate 3D information
methods like TLS or SfM so difficult. On the other hand, this difficulty is
on canopy structure possible. Key to this is the need to develop more
also the reason why indirect methods are so attractive. In conclusion,
robust, automated and flexible 3D canopy structure reconstruction
high quality validation data, tools and frameworks are needed to un­
methods. This is particularly true for very tall (> 60 m), dense tropical
derpin our ability to extract higher detail information from very accu­
forests and for TLS data with correspondingly greater variation in point
rate 3D data. There is ample opportunity to learn from other commu­
density from low to high in the canopy, laser beam spot size, and oc­
nities here who have used model intercomparison exercises to rapidly
clusion. Current 3D reconstruction approaches have generally been
develop fields such as radiative transfer modelling of the atmosphere
developed to work best on good quality (i.e. no noise and even point
and in vegetation (Widlowski et al., 2015) or filtering algorithms for
density) TLS point clouds, for single trees. This means that the more
digital elevation models from airborne laser scanning (Sithole and
time and resources that are put into data collection, registration, pre-
Vosselman, 2003).
filtering and then fine-tuning the reconstruction, the better the results
New ecological insight arising from 3D measurements is leading us
with these existing approaches. To advance cost-effective data collec­
to ask new and harder questions of our data. For example, TLS data can
tion, consensus needs to be reached over what constitutes ‘optimal’
be used to develop a structural economic spectrum to understand how
data; and how we objectively assess occlusion and point cloud quality.
woody plants arrange themselves along a few descriptive axis of
As we collect more data across a wider range of forested ecosystems,
structural traits (Verbeeck et al., 2019). TLS has increasingly been used
with different systems and under different conditions, the need for more
to support interdisciplinary research such as understanding habitat re­
rapid automated methods increases. 3D structure methods should ide­
quirements of mammals (Stobo-Wilson et al., 2020), monitoring but­
ally require little-to-no manual input and be agnostic to the specifics of
terfly populations (Hristov et al., 2019) or modelling leaf-deposition of
TLS system, data collection or ecosystem. A key additional benefit of
atmospheric particles in urban environments (Hofman et al., 2014,
fully-automated approaches is to drastically lower the barrier to entry
2016). Addressing new, often more complex, research questions leads
to the use of the data. Moreover, as 3D reconstruction approaches could
to greater appreciation of both the limitations but also the potential.
benefit from deep learning approaches (see Section 2.2), it should be
When we think about ways to improve the extraction of 3D canopy
possible to learn from every reconstruction so that each new branch,
structure information from TLS, two aspects should be considered. The
tree or forest reconstruction is approached with the inbuilt expertise of
first is time; high resolution temporal monitoring of 3D structure via
all previous reconstructions. We want to avoid constraining re­
repeated TLS scanning has the potential to deliver unique time-varying
constructions to exist in an echo chamber - ‘if you liked that tree, you'll
4D data of canopy structural dynamics. This is common-place in the
like this one’ - as that is the best way to filter out diversity and miss the
spectral domain, but arguably the structural domain provides as much,
unusual, which ecological systems have a habit of throwing up. This is
if not more, information. Developing ways to combine high time and
especially true for urban trees, which have a highly variable structural

11
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

space resolution 3D data with spectral information, will open up new the five anonymous reviewers for their valuable suggestions to improve
ways to analyse the relationships between structure and function, trait- the manuscript. We want to especially thank Alan Strahler for his lea­
based analysis and more. This will require new methods to enable dership within TLSIIG and TLSRCN and his invaluable input and
routine collection and merging of these data with special attention to mentoring in advancing TLS research.
quantifying occlusions and point cloud quality, but the rapid develop­
ment of TLS systems and methods is already heralding a new era in 4D References
measurement.
The second consideration is integration across spatial scales, as well Akerblom, M., Raumonen, P., Kaasalainen, M., Casella, E., 2015. Analysis of geometric
as with other measurements. UAV-LS is becoming increasingly primitives in quantitative structure models of tree stems. Remote Sens. 7, 4581–4603.
Åkerblom, M., Raumonen, P., Mäkipää, R., Kaasalainen, M., 2017. Automatic tree species
common, as is the development of low-range-low-cost LiDAR systems recognition with quantitative structure models. Remote Sens. Environ. 191, 1–12.
that could be deployed on mobile platforms and/or in large numbers Åkerblom, M., Raumonen, P., Casella, E., Disney, M.I., Danson, F.M., Gaulton, R.,
simultaneously within a forest system. We need to be thinking about Schofield, L.A., Kaasalainen, M., 2018. Non-intersecting leaf insertion algorithm for
tree structure models. Interface Focus 8, 20170045.
algorithms and data collection approaches that make the integration of Arel, I., Rose, D.C., Karnowski, T.P., 2010. Deep machine learning - a new frontier in
these data as seamless as possible. Various studies have already com­ artificial intelligence research [research frontier]. IEEE Comput. Intell. Mag. 5,
bined UAV, TLS, and ALS in one combination or another, but currently 13–18.
Ashcroft, M.B., Gollan, J.R., Ramp, D., 2014. Creating vegetation density profiles for a
these approaches are highly manually-intensive and typically require
diverse range of ecological habitats using terrestrial laser scanning. Methods Ecol.
downsampling or aggregation to some lower resolution or point density Evol. 5, 263–272.
than the native systems are capable of. This translates into new ques­ Asner, G.P., Scurlock, J.M.O., Hicke, A., 2003. Global synthesis of leaf area index ob­
servations: implications for ecological and remote sensing studies. Glob. Ecol.
tions as how do we avoid losing information like this and does it really
Biogeogr. 12, 191–205.
matter? We believe that currently there is no one-size-fits all approach, Atkins, J.W., Bohrer, G., Fahey, R.T., Hardiman, B.S., Morin, T.H., Stovall, A.E.L.,
and that searching for one is a mistake. We should accept that some Zimmerman, N., Gough, C.M., 2018a. Quantifying vegetation and canopy structural
applications will prioritise different data collection approaches (e.g. complexity from terrestrial LiDAR data using the forestr r package. Methods Ecol.
Evol. 9, 2057–2066.
single scans vs high density scanning) and require different levels of Atkins, J.W., Fahey, R.T., Hardiman, B.H., Gough, C.M., 2018b. Forest canopy structural
detail of the derived structural parameters (e.g. total tree AGB vs. complexity and light absorption relationships at the subcontinental scale. J. Geophys.
diameter and angles of higher order branches). Another aspect to Res. Biogeosci. 123, 1387–1405.
Baines, O., Wilkes, P., Disney, M., 2020. Quantifying urban forest structure with open-
consider is the inclusion of other data sources or data fusion. For ex­ access remote sensing data sets. Urban For. Urban Green. 50, 126653.
ample, SfM is developing as rapidly as TLS methods, if not more so. Balsi, M., Esposito, S., Fallavollita, P., Nardinocchi, C., 2018. Single-tree detection in
New TLS approaches ought to be very open to integration with SfM to high-density LiDAR data from UAV-based survey. Eur. J. Remote Sens. 51, 679–692.
Barbeito, I., Dassot, M., Bayer, D., Collet, C., Drössler, L., Löf, M., del Rio, M., Ruiz-
make the most of all the new tools at our disposal. This requires Peinado, R., Forrester, D.I., Bravo-Oviedo, A., Pretzsch, H., 2017. Terrestrial laser
thinking about tools and software that will make this possible, which in scanning reveals differences in crown structure of Fagus sylvatica in mixed vs. pure
turn needs ecologists with training in these methods and how to de­ European forests. For. Ecol. Manag. 405, 381–390.
Bastin, J.-F., Fayolle, A., Tarelkin, Y., Van den Bulcke, J., de Haulleville, T., Mortier, F.,
velop them.
Beeckman, H., Van Acker, J., Serckx, A., Bogaert, J., De Cannière, C., 2015. Wood
One way to potentially accelerate this development is by bringing in specific gravity variations and biomass of Central African tree species: the simple
expertise from across domains. If we can define our problems in forest choice of the outer wood. PLoS One 10, e0142146.
Bauwens, S., Bartholomeus, H., Calders, K., Lejeune, P., 2016. Forest inventory with
ecology more generally, so that they are attractive for other commu­
terrestrial LiDAR: a comparison of static and hand-held mobile laser scanning. Forests
nities (e.g. remote sensing, physics, engineering, computer science and 7, 127.
ML etc.) then there is clearly the potential for more rapid advancement Bayer, D., Seifert, S., Pretzsch, H., 2013. Structural crown properties of Norway spruce
than solely by trying to turn forest ecologists into computer scientists. (Picea abies [L.] Karst.) and European beech (Fagus sylvatica [L.]) in mixed versus
pure stands revealed by terrestrial laser scanning. Trees 27, 1035–1047.
These are issues that will be central to the next generation of ‘4D Béland, M., Widlowski, J.-L., Fournier, R.A., Côté, J.-F., Verstraete, M.M., 2011.
ecologists’ and the insights they will provide in forest ecology. Estimating leaf area distribution in savanna trees from terrestrial LiDAR measure­
ments. Agric. For. Meteorol. 151, 1252–1266.
Béland, M., Baldocchi, D.D., Widlowski, J.-L., Fournier, R.A., Verstraete, M.M., 2014a. On
Declaration of Competing Interest seeing the wood from the leaves and the role of voxel size in determining leaf area
distribution of forests with terrestrial LiDAR. Agric. For. Meteorol. 184, 82–97.
Béland, M., Widlowski, J.-L., Fournier, R.A., 2014b. A model for deriving voxel-level tree
The authors declare that they have no known competing financial leaf area density estimates from ground-based LiDAR. Environ. Model. Softw. 51,
interests or personal relationships that could have appeared to influ­ 184–189.
Beland, M., Parker, G., Sparrow, B., Harding, D., Chasmer, L., Phinn, S., Antonarakis, A.,
ence the work reported in this paper. Strahler, A., 2019. On promoting the use of lidar systems in forest ecosystem re­
search. For. Ecol. Manag. 450, 117484.
Belton, D., Moncrieff, S., Chapman, J., 2013. Processing tree point clouds using Gaussian
Acknowledgements mixture models. ISPRS annals of photogrammetry. Remote Sens. Spatial Informa. Sci.
II-5 (W2), 43–48.
Bentley, L.P., Stegen, J.C., Savage, V.M., Smith, D.D., von Allmen, E.I., Sperry, J.S., Reich,
This paper is the result of the 4th scientific meeting on terrestrial
P.B., Enquist, B.J., 2013. An empirical assessment of tree branching networks and
laser scanning in forest ecology hosted by Ghent University in 2019. implications for plant allometric scaling models. Ecol. Lett. 16, 1069–1078.
Travel funding for some US participants was provided by the NSF Bienert, A., Georgi, L., Kunz, M., Maas, H.-G., Von Oheimb, G., 2018. Comparison and
Terrestrial Laser Scanning (TLS)RCN 1455636 (RCN-IBDR: combination of Mobile and terrestrial laser scanning for natural Forest inventories.
Forests 9, 395.
Coordinating the Development of Terrestrial Lidar Scanning for Blair, J.B., Hofton, M.A., 1999. Modeling laser altimeter return waveforms over complex
Aboveground Biomass and Ecological Applications). KC was funded by vegetation using high-resolution elevation data. Geophys. Res. Lett. 2509–2512.
the European Union's Horizon 2020 research and innovation pro­ Boni Vicari, M., Disney, M., Wilkes, P., Burt, A., Calders, K., Woodgate, W., 2019a. Leaf
and wood classification framework for terrestrial LiDAR point clouds. Methods Ecol.
gramme under the Marie Sklodowska-Curie grant agreement No Evol. 10, 680–694.
835398. HV and SMSM were funded by BELSPO (Belgian Science Policy Boni Vicari, M., Pisek, J., Disney, M., 2019b. New estimates of leaf angle distribution from
Office) in the frame of the STEREO III programme – project 3D-FOREST terrestrial LiDAR: comparison with measured and modelled estimates from nine
broadleaf tree species. Agric. For. Meteorol. 264, 322–333.
(SR/02/355). NS was funded by the Academy of Finland (project Boucher, P., 2019. Characterizing the Impacts of the Invasive Hemlock Woolly Adelgid on
number 315079). JC acknowledges funding from “Programme the Forest Structure of New England. Univ. Massachusetts, Boston.
Investissement d'Avenir” managed by Agence Nationale de la Brede, B., Lau, A., Bartholomeus, H.M., Kooistra, L., 2017. Comparing RIEGL RiCOPTER
UAV LiDAR derived canopy height and DBH with terrestrial LiDAR. Sensors 17.
Recherche (ANR-10-LABX-25-01; ANR-10-LABX-0041). The Litchfield Brede, B., Calders, K., Lau, A., Raumonen, P., Bartholomeus, H.M., Herold, M., Kooistra,
ALS data (Fig. 5) was provided by the TERN AusCover data facility. The L., 2019. Non-destructive tree volume estimation through quantitative structure
authors thank Peter Boucher and Laura Duncanson for their input, and modelling: comparing UAV laser scanning with terrestrial LIDAR. Remote Sens.

12
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

Environ. 233, 111355. Disney, M., Burt, A., Calders, K., Schaaf, C., Stovall, A., 2019. Innovations in ground and
Bucksch, A., Lindenbergh, R., 2008. CAMPINO—A skeletonization method for point cloud airborne technologies as reference and for training and validation: terrestrial laser
processing. ISPRS J. Photogramm. Remote Sens. 63, 115–127. scanning (TLS). Surv. Geophys. 1–22.
Burt, A., Disney, M., Calders, K., 2019. Extracting individual trees from lidar point clouds Donager, J.J., Sankey, T.T., Sankey, J.B., Sanchez Meador, A.J., Springer, A.E., Bailey,
using treeseg. Methods Ecol. Evol. 10, 438–445. J.D., 2018. Examining forest structure with terrestrial Lidar: suggestions and novel
Burt, A., Calders, K., Cuni-Sanchez, A., Gómez-Dans, J., Lewis, P., Lewis, S.L., Malhi, Y., techniques based on comparisons between scanners and Forest treatments. Life
Phillips, O.L., Disney, M., 2020. Assessment of bias in pan-tropical biomass predic­ Support Biosph. Sci. 5, 753–776.
tions. Front. For. Glob. Change 3, 12. Douglas, E.S., Martel, J., Li, Z., Howe, G., Hewawasam, K., Marshall, R.A., Schaaf, C.L.,
Cabo, C., Del Pozo, S., Rodríguez-Gonzálvez, P., Ordóñez, C., González-Aguilera, D., Cook, T.A., Newnham, G.J., Strahler, A., Chakrabarti, S., 2015. Finding leaves in the
2018. Comparing terrestrial laser scanning (TLS) and wearable laser scanning (WLS) forest: the dual-wavelength Echidna Lidar. IEEE Geosci. Remote Sens. Lett. 12,
for individual tree modeling at plot level. Remote Sens. 10, 540. 776–780.
Calders, K., Armston, J., Newnham, G., Herold, M., Goodwin, N., 2014. Implications of Dubayah, R., Blair, J.B., Goetz, S., Fatoyinbo, L., Hansen, M., Healey, S., Hofton, M.,
sensor configuration and topography on vertical plant profiles derived from terres­ Hurtt, G., Kellner, J., Luthcke, S., Armston, J., Tang, H., Duncanson, L., Hancock, S.,
trial LiDAR. Agric. For. Meteorol. 194, 104–117. Jantz, P., Marselis, S., Patterson, P.L., Qi, W., Silva, C., 2020. The global ecosystem
Calders, K., Newnham, G., Burt, A., Murphy, S., Raumonen, P., Herold, M., Culvenor, D., dynamics investigation: high-resolution laser ranging of the earth’s forests and to­
Avitabile, V., Disney, M., Armston, J., et al., 2015a. Nondestructive estimates of pography. Sci. Remote Sens. 1, 100002.
above-ground biomass using terrestrial laser scanning. Methods Ecol. Evol. 6, Duncanson, L., Armston, J., Disney, M., Avitabile, V., Barbier, N., Calders, K., Carter, S.,
198–208. Chave, J., Herold, M., Crowther, T.W., Falkowski, M., Kellner, J.R., Labrière, N.,
Calders, K., Schenkels, T., Bartholomeus, H., Armston, J., Verbesselt, J., Herold, M., Lucas, R., MacBean, N., McRoberts, R.E., Meyer, V., Næsset, E., Nickeson, J.E., Paul,
2015b. Monitoring spring phenology with high temporal resolution terrestrial LiDAR K.I., Phillips, O.L., Réjou-Méchain, M., Román, M., Roxburgh, S., Saatchi, S.,
measurements. Agric. For. Meteorol. 203, 158–168. Schepaschenko, D., Scipal, K., Siqueira, P.R., Whitehurst, A., Williams, M., 2019. The
Calders, K., Disney, M.I., Armston, J., Burt, A., Brede, B., Origo, N., Muir, J., Nightingale, importance of consistent global Forest aboveground biomass product validation.
J., 2017. Evaluation of the range accuracy and the radiometric calibration of multiple Surv. Geophys. 40, 979–999.
terrestrial laser scanning instruments for data interoperability. IEEE Trans. Geosci. Eitel, J.U.H., Vierling, L.A., Long, D.S., 2010. Simultaneous measurements of plant
Remote Sens. 55, 2716–2724. structure and chlorophyll content in broadleaf saplings with a terrestrial laser
Calders, K., Origo, N., Burt, A., Disney, M., Nightingale, J., Raumonen, P., Åkerblom, M., scanner. Remote Sens. Environ. 114, 2229–2237.
Malhi, Y., Lewis, P., 2018a. Realistic forest stand reconstruction from terrestrial Elsherif, A., Gaulton, R., Mills, J., 2019a. Four dimensional mapping of vegetation
LiDAR for radiative transfer modelling. Remote Sens. 10, 933. moisture content using dual-wavelength terrestrial laser scanning. Remote Sens. 11,
Calders, K., Origo, N., Disney, M., Nightingale, J., Woodgate, W., Armston, J., Lewis, P., 2311.
2018b. Variability and bias in active and passive ground-based measurements of Elsherif, A., Gaulton, R., Shenkin, A., Malhi, Y., Mills, J., 2019b. Three dimensional
effective plant, wood and leaf area index. Agric. For. Meteorol. 252, 231–240. mapping of forest canopy equivalent water thickness using dual-wavelength terres­
Calders, K., Phinn, S., Ferrari, R., Leon, J., Armston, J., Asner, G.P., Disney, M., 2020. 3D trial laser scanning. Agric. For. Meteorol. 276-277, 107627.
imaging insights into forests and coral reefs. Trends Ecol. Evol. 6–9. Enquist, B.J., Kerkhoff, A.J., Stark, S.C., Swenson, N.G., McCarthy, M.C., Price, C.A.,
Case, B.S., Hall, R.J., 2008. Assessing prediction errors of generalized tree biomass and 2007. A general integrative model for scaling plant growth, carbon flux, and func­
volume equations for the boreal forest region of west-Central Canada. Can. J. For. tional trait spectra. Nature 449, 218–222.
Res. 38, 878–889. Enquist, B.J., West, G.B., Brown, J.H., 2009. Extensions and evaluations of a general
Chave, J., Andalo, C., Brown, S., Cairns, M.A., Chambers, J.Q., Eamus, D., Fölster, H., quantitative theory of forest structure and dynamics. Proc. Natl. Acad. Sci. U. S. A.
Fromard, F., Higuchi, N., Kira, T., Lescure, J.-P., Nelson, B.W., Ogawa, H., Puig, H., 106, 7046–7051.
Riéra, B., Yamakura, T., 2005. Tree allometry and improved estimation of carbon Eriksson, E., 2006. Thinning operations and their impact on biomass production in stands
stocks and balance in tropical forests. Oecologia 145, 87–99. of Norway spruce and scots pine. Biomass Bioenergy 30, 848–854.
Chave, J., Réjou-Méchain, M., Búrquez, A., Chidumayo, E., Colgan, M.S., Delitti, W.B.C., Fernández-Sarría, A., López-Cortés, I., Estornell, J., Velázquez-Martí, B., Salazar, D.,
Duque, A., Eid, T., Fearnside, P.M., Goodman, R.C., Henry, M., Martínez-Yrízar, A., 2019. Estimating residual biomass of olive tree crops using terrestrial laser scanning.
Mugasha, W.A., Muller-Landau, H.C., Mencuccini, M., Nelson, B.W., Ngomanda, A., Int. J. Appl. Earth Obs. Geoinf. 75, 163–170.
Nogueira, E.M., Ortiz-Malavassi, E., Pélissier, R., Ploton, P., Ryan, C.M., Saldarriaga, Fischer, M., Huss, M., Kummert, M., Hoelzle, M., 2016. Application and validation of
J.G., Vieilledent, G., 2014. Improved allometric models to estimate the aboveground long-range terrestrial laser scanning to monitor the mass balance of very small gla­
biomass of tropical trees. Glob. Chang. Biol. 20, 3177–3190. ciers in the Swiss Alps. Cryosphere 10, 1279–1295.
Chave, J., Davies, S.J., Phillips, O.L., Lewis, S.L., Sist, P., Schepaschenko, D., Armston, J., Fleck, S., Mölder, I., Jacob, M., Gebauer, T., Jungkunst, H.F., Leuschner, C., 2011.
Baker, T.R., Coomes, D., Disney, M., Duncanson, L., Hérault, B., Labrière, N., Meyer, Comparison of conventional eight-point crown projections with LIDAR-based virtual
V., Réjou-Méchain, M., Scipal, K., Saatchi, S., 2019. Ground data are essential for crown projections in a temperate old-growth forest. Ann. For. Sci. 68, 1173–1185.
biomass remote sensing missions. Surv. Geophys. 40, 863–880. Fyllas, N.M., Bentley, L.P., Shenkin, A., Asner, G.P., Atkin, O.K., Díaz, S., Enquist, B.J.,
Chen, Y., Zhu, X., Yebra, M., Harris, S., Tapper, N., 2016. Strata-based forest fuel clas­ Farfan-Rios, W., Gloor, E., Guerrieri, R., Huasco, W.H., Ishida, Y., Martin, R.E., Meir,
sification for wild fire hazard assessment using terrestrial LiDAR. J. Appl. Remote. P., Phillips, O., Salinas, N., Silman, M., Weerasinghe, L.K., Zaragoza-Castells, J.,
Sens. 10, 046025. Malhi, Y., 2017. Solar radiation and functional traits explain the decline of forest
Chen, Y., Zhu, X., Yebra, M., Harris, S., Tapper, N., 2017. Development of a predictive primary productivity along a tropical elevation gradient. Ecol. Lett. 20, 730–740.
model for estimating forest surface fuel load in Australian eucalypt forests with Gabbud, C., Micheletti, N., Lane, S.N., 2015. Lidar measurement of surface melt for a
LiDAR data. Environ. Model. Softw. 97, 61–71. temperate Alpine glacier at the seasonal and hourly scales. J. Glaciol. 61, 963–974.
Cifuentes, R., Van der Zande, D., Salas-Eljatib, C., Farifteh, J., Coppin, P., 2018. A si­ Gaulton, R., Danson, F.M., Ramirez, F.A., Gunawan, O., 2013. The potential of dual-
mulation study using terrestrial LiDAR point cloud data to quantify spectral varia­ wavelength laser scanning for estimating vegetation moisture content. Remote Sens.
bility of a broad-leaved Forest canopy. Sensors 18, 3357. Environ. 132, 32–39.
Côté, J.-F., Widlowski, J.-L., Fournier, R.A., Verstraete, M.M., 2009. The structural and Georgi, L., Kunz, M., Fichtner, A., Härdtle, W., Reich, K.F., Sturm, K., Welle, T., von
radiative consistency of three-dimensional tree reconstructions from terrestrial lidar. Oheimb, G., 2018. Long-term abandonment of Forest management has a strong im­
Remote Sens. Environ. 113, 1067–1081. pact on tree morphology and wood volume allocation pattern of European beech
Culvenor, D.S., Newnham, G.J., Mellor, A., Sims, N.C., Haywood, A., 2014. Automated in- (Fagus sylvatica L.). Forests 9, 704.
situ laser scanner for monitoring forest leaf area index. Sensors 14, 14994–15008. Gielen, B., Acosta, M., Altimir, N., Buchmann, N., Cescatti, A., Ceschia, E., Fleck, S.,
Cuni-Sanchez, A., White, L.J.T., Calders, K., Jeffery, K.J., Abernethy, K., Burt, A., Disney, Hörtnagl, L., Klumpp, K., Kolari, P., Lohila, A., Loustau, D., Marañón-Jiménez, S.,
M., Gilpin, M., Gomez-Dans, J.L., Lewis, S.L., 2016. African savanna-Forest boundary Manise, T., Matteucci, G., Merbold, L., Herschlein, C., Moureaux, C., Montagnani, L.,
dynamics: a 20-year study. PLoS One 11, e0156934. Wohlfahrt, G., 2018. Ancillary vegetation measurements at ICOS ecosystem stations.
Danson, F.M., Gaulton, R., Armitage, R.P., Disney, M., Gunawan, O., Lewis, P., Pearson, Int. Agrophys. 32, 645–664.
G., Ramirez, F.A., 2014. Developing a dual-wavelength full-waveform terrestrial laser Gonzalez de Tanago, J., Lau, A., Bartholomeus, H., Herold, M., Avitabile, V., Raumonen,
scanner to characterize forest canopy structure. Agric. For. Meteorol. 198-199, 7–14. P., Martius, C., Goodman, R.C., Disney, M., Manuri, S., Burt, A., Calders, K., 2018.
Danson, M.F., Sasse, F., Schofield, L.A., 2018. Spectral and spatial information from a Estimation of above-ground biomass of large tropical trees with terrestrial LiDAR.
novel dual-wavelength full-waveform terrestrial laser scanner for forest ecology. Methods Ecol. Evol. 9, 223–234.
Interface Focus 8, 20170049. Goodfellow, I., Bengio, Y., Courville, A., 2016. Deep Learning. MIT Press.
Decuyper, M., Mulatu, K.A., Brede, B., Calders, K., Armston, J., Rozendaal, D.M.A., Mora, Goodman, R.C., Phillips, O.L., Baker, T.R., 2014. The importance of crown dimensions to
B., Clevers, J.G., Kooistra, L., Herold, M., et al., 2018. Assessing the structural dif­ improve tropical tree biomass estimates. Ecol. Appl. 24, 680–698.
ferences between tropical forest types using terrestrial laser scanning. For. Ecol. Gottfried, M., Hollaus, M., Glira, P., Wieser, M., Milenković, M., Riegl, U., Pfennigbauer,
Manag. 429, 327–335. M., 2015. First examples from the RIEGL VUX-SYS for forestry applications. Proceed.
Disney, M., 2016. Remote sensing of vegetation: Potentials, limitations, developments SilviLaser 2015, 105–107.
and applications. In: Hikosaka, K., Niinemets, Ü., Anten, N.P.R. (Eds.), Canopy Grau, E., Durrieu, S., Fournier, R., Gastellu-Etchegorry, J.-P., Yin, T., 2017. Estimation of
Photosynthesis: From Basics to Applications. Springer Netherlands, Dordrecht, pp. 3D vegetation density with terrestrial laser scanning data using voxels. A sensitivity
289–331. analysis of influencing parameters. Remote Sens. Environ. 191, 373–388.
Disney, M., 2019. Terrestrial LiDAR: a three-dimensional revolution in how we look at Guo, Q., Su, Y., Hu, T., Zhao, X., Wu, F., Li, Y., Liu, J., Chen, L., Xu, G., Lin, G., Zheng, Y.,
trees. New Phytol. 222, 1736–1741. Lin, Y., Mi, X., Fei, L., Wang, X., 2017. An integrated UAV-borne lidar system for 3D
Disney, M.I., Boni Vicari, M., Burt, A., Calders, K., Lewis, S.L., Raumonen, P., Wilkes, P., habitat mapping in three forest ecosystems across China. Int. J. Remote Sens. 38,
2018. Weighing trees with lasers: advances, challenges and opportunities. Interface 2954–2972.
Focus 8, 20170048. Hackenberg, J., Spiecker, H., Calders, K., Disney, M., Raumonen, P., 2015.

13
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

SimpleTree—an efficient open source tool to build tree models from TLS clouds. laser called Echidna. Tree Physiol. 29, 171–181.
Forests 6, 4245–4294. Kaasalainen, S., Krooks, A., Liski, J., Raumonen, P., Kaartinen, H., Kaasalainen, M.,
Hajek, P., Seidel, D., Leuschner, C., 2015. Mechanical abrasion, and not competition for Puttonen, E., Anttila, K., Mäkipää, R., 2014. Change detection of tree biomass with
light, is the dominant canopy interaction in a temperate mixed forest. For. Ecol. terrestrial laser scanning and quantitative structure modelling. Remote Sens. 6,
Manag. 348, 108–116. 3906–3922.
Hakala, T., Suomalainen, J., Kaasalainen, S., Chen, Y., 2012. Full waveform hyperspectral Kahlmann, T., Remondino, F., Ingensand, H., 2006. Calibration for increased accuracy of
LiDAR for terrestrial laser scanning. Opt. Express 20, 7119–7127. the range imaging camera swissranger. In: Proceedings of the ISPRS Commission V
Hallé, F., Oldeman, R.A.A., Tomlinson, P.B., 1978. Opportunistic tree architecture. Symposium’Image Engineering and Vision Metrology'. Isprs, pp. 136–141.
Trop.Trees For. 269–331. Kankare, V., Holopainen, M., Vastaranta, M., Puttonen, E., Yu, X., Hyyppä, J., Vaaja, M.,
Halupka, K., Garnavi, R., Moore, S., 2019. Deep semantic instance segmentation of tree- Hyyppä, H., Alho, P., 2013. Individual tree biomass estimation using terrestrial laser
like structures using synthetic data. In: 2019 IEEE Winter Conference on Applications scanning. ISPRS J. Photogramm. Remote Sens. 75, 64–75.
of Computer Vision (WACV), pp. 1713–1722. Kao, R.H., Gibson, C.M., Gallery, Rachel E., Meier, C.L., Barnett, D.T., Docherty, K.M.,
Hancock, S., Essery, R., Reid, T., Carle, J., Baxter, R., Rutter, N., Huntley, B., 2014. Blevins, K.K., Travers, P.D., Azuaje, E., Springer, Y.P., Thibault, K.M., McKenzie, V.J.,
Characterising forest gap fraction with terrestrial lidar and photography: an ex­ Keller, M., Alves, L.F., Hinckley, E.-L.S., Parnell, J., Schimel, D., 2012. NEON ter­
amination of relative limitations. Agric. For. Meteorol. 189-190, 105–114. restrial field observations: designing continental-scale, standardized sampling.
Hancock, S., Anderson, K., Disney, M., Gaston, K.J., 2017. Measurement of fine-spatial- Ecosphere 3, 115.
resolution 3D vegetation structure with airborne waveform lidar: calibration and Karan, M., Liddell, M., Prober, S.M., Arndt, S., Beringer, J., Boer, M., Cleverly, J., Eamus,
validation with voxelised terrestrial lidar. Remote Sens. Environ. 188, 37–50. D., Grace, P., Van Gorsel, E., Hero, J.-M., Hutley, L., Macfarlane, C., Metcalfe, D.,
Hancock, S., Armston, J., Hofton, M., Sun, X., Tang, H., Duncanson, L.I., Kellner, J.R., Meyer, W., Pendall, E., Sebastian, A., Wardlaw, T., 2016. The Australian supersite
Dubayah, R., 2019. The GEDI simulator: a large-footprint waveform lidar simulator network: a continental, long-term terrestrial ecosystem observatory. Sci. Total
for calibration and validation of spaceborne missions. Earth Space Sci. 294–310. Environ. 568, 1263–1274.
Hartzell, P.J., Glennie, C.L., Finnegan, D.C., 2015. Empirical waveform decomposition Kelbe, D., van Aardt, J., Romanczyk, P., van Leeuwen, M., Cawse-Nicholson, K., 2016.
and radiometric calibration of a terrestrial full-waveform laser scanner. IEEE Trans. Marker-free registration of forest terrestrial laser scanner data pairs with embedded
Geosci. Remote Sens. 53, 162–172. confidence metrics. IEEE Trans. Geosci. Remote Sens. 54, 4314–4330.
Henning, J.G., Radtke, P.J., 2006. Detailed stem measurements of standing trees from Kellner, J.R., Armston, J., Birrer, M., Cushman, K.C., Duncanson, L., Eck, C., Falleger, C.,
ground-based scanning Lidar. For. Sci. 52, 67–80. Imbach, B., Král, K., Krůček, M., Trochta, J., Vrška, T., Zgraggen, C., 2019. New
Henry, M., Picard, N., Trotta, C., Manlay, R., Valentini, R., Bernoux, M., Saint-André, L., opportunities for forest remote sensing through ultra-high-density drone Lidar. Surv.
2011. Estimating tree biomass of sub-Saharan African forests: a review of available Geophys. 40, 959–977.
allometric equations. Silva Fenn. 45, 38. Krishna Moorthy, S.M., Bao, Y., Calders, K., Schnitzer, S.A., Verbeeck, H., 2019a. Semi-
Hess, C., Härdtle, W., Kunz, M., Fichtner, A., von Oheimb, G., 2018. A high-resolution automatic extraction of liana stems from terrestrial LiDAR point clouds of tropical
approach for the spatiotemporal analysis of forest canopy space using terrestrial laser rainforests. ISPRS J. Photogramm. Remote Sens. 154, 114–126.
scanning data. Ecol. Evol. 8, 6800–6811. Krishna Moorthy, S.M., Calders, K., Boni Vicari, M., Verbeeck, H., 2019b. Improved su­
Hofman, J., Bartholomeus, H., Calders, K., Van Wittenberghe, S., Wuyts, K., Samson, R., pervised learning-based approach for leaf and wood classification From LiDAR point
2014. On the relation between tree crown morphology and particulate matter de­ clouds of forests. IEEE Trans. Geosci. Remote Sens. 1–14.
position on urban tree leaves: a ground-based LiDAR approach. Atmos. Environ. 99, Krishna Moorthy, S.M., Raumonen, P., Van den Bulcke, J., Calders, K., Verbeeck, H.,
130–139. 2020. Terrestrial laser scanning for non-destructive estimates of liana stem biomass.
Hofman, J., Bartholomeus, H., Janssen, S., Calders, K., Wuyts, K., Van Wittenberghe, S., For. Ecol. Manag. 456, 117751.
Samson, R., 2016. Influence of tree crown characteristics on the local PM10 dis­ Krůček, M., Trochta, J., Cibulka, M., Král, K., 2019. Beyond the cones: how crown shape
tribution inside an urban street canyon in Antwerp (Belgium): a model and experi­ plasticity alters aboveground competition for space and light—evidence from ter­
mental approach. Urban For. Urban Green. 20, 265–276. restrial laser scanning. Agric. For. Meteorol. 264, 188–199.
Hopkinson, C., Chasmer, L., Young-Pow, C., Treitz, P., 2004. Assessing forest metrics with Kunz, M., Fichtner, A., Härdtle, W., Raumonen, P., Bruelheide, H., von Oheimb, G., 2019.
a ground-based scanning lidar. Can. J. For. Res. 34, 573–583. Neighbour species richness and local structural variability modulate aboveground
Hosoi, F., Omasa, K., 2006. Voxel-based 3-D modeling of individual trees for estimating allocation patterns and crown morphology of individual trees. Ecol. Lett. 22,
leaf area density using high-resolution portable scanning Lidar. IEEE Trans. Geosci. 2130–2140.
Remote Sens. 44, 3610–3618. Kuusk, A., 2020. Leaf orientation measurement in a mixed hemiboreal broadleaf forest
Hristov, N.I., Nikolaidis, D., Hubel, T.Y., Allen, L.C., 2019. Estimating overwintering stand using terrestrial laser scanner. Trees 34, 371–380.
monarch butterfly populations using terrestrial LiDAR scanning. Front. Ecol. Evol. 7, Lau, A., Bentley, L.P., Martius, C., Shenkin, A., Bartholomeus, H., Raumonen, P., Malhi,
266. Y., Jackson, T., Herold, M., 2018. Quantifying branch architecture of tropical trees
Hudak, A.T., Evans, J.S., Stuart Smith, A.M., 2009. LiDAR utility for natural resource using terrestrial LiDAR and 3D modelling. Trees 32, 1219–1231.
managers. Remote Sens. 1, 934–951. Lau, A., Calders, K., Bartholomeus, H., Martius, C., Raumonen, P., Herold, M., Vicari, M.,
Huo, L., Zhang, N., Zhang, X., Wu, Y., 2019. Tree defoliation classification based on point Sukhdeo, H., Singh, J., Goodman, R.C., 2019a. Tree biomass equations from terres­
distribution features derived from single-scan terrestrial laser scanning data. Ecol. trial LiDAR: a case study in Guyana. Forests 10, 527.
Indic. 103, 782–790. Lau, A., Martius, C., Bartholomeus, H., Shenkin, A., Jackson, T., Malhi, Y., Herold, M.,
Hyyppä, E., Kukko, A., Kaijaluoto, R., White, J.C., Wulder, M.A., Pyörälä, J., Liang, X., Yu, Bentley, L.P., 2019b. Estimating architecture-based metabolic scaling exponents of
X., Wang, Y., Kaartinen, H., Virtanen, J.-P., Hyyppä, J., 2020. Accurate derivation of tropical trees using terrestrial LiDAR and 3D modelling. For. Ecol. Manag. 439,
stem curve and volume using backpack mobile laser scanning. ISPRS J. Photogramm. 132–145.
Remote Sens. 161, 246–262. Lefsky, M.A., McHale, M.R., 2008. Volume estimates of trees with complex architecture
Itakura, K., Hosoi, F., 2019. Estimation of leaf inclination angle in three-dimensional from terrestrial laser scanning. J. Appl. Remote. Sens. 2, 023521.
plant images obtained from Lidar. Remote Sens. 11, 344. Lercari, N., 2016. Terrestrial laser scanning in the age of sensing. In: Forte, M., Campana,
Jaakkola, A., Hyyppä, J., Kukko, A., Yu, X., Kaartinen, H., Lehtomäki, M., Lin, Y., 2010. A S. (Eds.), Digital Methods and Remote Sensing in Archaeology: Archaeology in the
low-cost multi-sensoral mobile mapping system and its feasibility for tree measure­ Age of Sensing. Springer International Publishing, Cham, pp. 3–33.
ments. ISPRS J. Photogramm. Remote Sens. 65, 514–522. Li, Z., Jupp, D.L.B., Strahler, A.H., Schaaf, C.B., Howe, G., Hewawasam, K., Douglas, E.S.,
Jackson, T., Shenkin, A., Moore, J., Bunce, A., van Emmerik, T., Kane, B., Burcham, D., Chakrabarti, S., Cook, T.A., Paynter, I., Saenz, E.J., Schaefer, M., 2016. Radiometric
James, K., Selker, J., Calders, K., Origo, N., Disney, M., Burt, A., Wilkes, P., calibration of a dual-wavelength, full-waveform terrestrial Lidar. Sensors 16, 313.
Raumonen, P., Gonzalez de Tanago Menaca, J., Lau, A., Herold, M., Goodman, R.C., Li, Z., Schaefer, M., Strahler, A., Schaaf, C., Jupp, D., 2018. On the utilization of novel
Fourcaud, T., Malhi, Y., 2019. An architectural understanding of natural sway fre­ spectral laser scanning for three-dimensional classification of vegetation elements.
quencies in trees. J. R. Soc. Interface 16, 20190116. Interface Focus 8, 20170039.
Jonckheere, I., Fleck, S., Nackaerts, K., Muys, B., Coppin, P., Weiss, M., Baret, F., 2004. Li, J., Wang, N., Zhang, L., Du, B., Tao, D., 2020. Recurrent feature reasoning for image
Review of methods for in situ leaf area index determination: part I. Theories, sensors inpainting. In: Proceedings of the IEEE/CVF Conference on Computer Vision and
and hemispherical photography. Agric. For. Meteorol. 121, 19–35. Pattern Recognition, pp. 7760–7768.
Juchheim, J., Ammer, C., Schall, P., Seidel, D., 2017a. Canopy space filling rather than Liang, X., Litkey, P., Hyyppa, J., Kaartinen, H., Vastaranta, M., Holopainen, M., 2012.
conventional measures of structural diversity explains productivity of beech stands. Automatic stem mapping using single-scan terrestrial laser scanning. IEEE Trans.
For. Ecol. Manag. 395, 19–26. Geosci. Remote Sens. 50, 661–670.
Juchheim, J., Annighöfer, P., Ammer, C., Calders, K., Raumonen, P., Seidel, D., 2017b. Liang, X., Wang, Y., Jaakkola, A., Kukko, A., Kaartinen, H., Hyyppä, J., Honkavaara, E.,
How management intensity and neighborhood composition affect the structure of Liu, J., 2015. Forest data collection using terrestrial image-based point clouds from a
beech (Fagus sylvatica L.) trees. Trees 31, 1723–1735. handheld camera compared to terrestrial and personal laser scanning. IEEE Trans.
Juchheim, J., Ehbrecht, M., Schall, P., Ammer, C., Seidel, D., 2020. Effect of tree species Geosci. Remote Sens. 53, 5117–5132.
mixing on stand structural complexity. Forestry 93, 75–83. Liang, X., Kankare, V., Hyyppä, J., Wang, Y., Kukko, A., Haggrén, H., Yu, X., Kaartinen,
Junttila, S., Holopainen, M., Vastaranta, M., Lyytikäinen-Saarenmaa, P., Kaartinen, H., H., Jaakkola, A., Guan, F., Holopainen, M., Vastaranta, M., 2016. Terrestrial laser
Hyyppä, J., Hyyppä, H., 2019. The potential of dual-wavelength terrestrial lidar in scanning in forest inventories. ISPRS J. Photogramm. Remote Sens. 115, 63–77.
early detection of Ips typographus (L.) infestation – leaf water content as a proxy. Liang, X., Hyyppä, J., Kaartinen, H., Lehtomäki, M., Pyörälä, J., Pfeifer, N., Holopainen,
Remote Sens. Environ. 231, 111264. M., Brolly, G., Francesco, P., Hackenberg, J., Huang, H., Jo, H.-W., Katoh, M., Liu, L.,
Juodvalkis, A., Kairiukstis, L., Vasiliauskas, R., 2005. Effects of thinning on growth of six Mokroš, M., Morel, J., Olofsson, K., Poveda-Lopez, J., Trochta, J., Wang, D., Wang, J.,
tree species in north-temperate forests of Lithuania. Eur. J. For. Res. 124, 187–192. Xi, Z., Yang, B., Zheng, G., Kankare, V., Luoma, V., Yu, X., Chen, L., Vastaranta, M.,
Jupp, D.L.B., Culvenor, D.S., Lovell, J.L., Newnham, G.J., Strahler, A.H., Woodcock, C.E., Saarinen, N., Wang, Y., 2018. International benchmarking of terrestrial laser scan­
2009. Estimating forest LAI profiles and structural parameters using a ground-based ning approaches for forest inventories. ISPRS J. Photogramm. Remote Sens. 144,

14
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

137–179. Effects of Different Thinning Programmes on Stand Level Gross- and Net Stem
Liang, X., Wang, Y., Pyörälä, J., Lehtomäki, M., Yu, X., Kaartinen, H., Kukko, A., Volume Production. Studia forestalia Suecica. Swedish University of Agricultural
Honkavaara, E., Issaoui, A.E.I., Nevalainen, O., Vaaja, M., Virtanen, J.-P., Katoh, M., Sciences, Faculty of Forest Sciences, Umeå.
Deng, S., 2019. Forest in situ observations using unmanned aerial vehicle as an al­ Othmani, A., Lew Yan Voon, L.F.C., Stolz, C., Piboule, A., 2013. Single tree species
ternative of terrestrial measurements. For. Ecosyst. 6, 20. classification from terrestrial laser scanning data for forest inventory. Pattern Recogn.
Lintermann, B., Deussen, O., 1999. Interactive modelling of plants. IEEE Comput. Graph. Lett. 34, 2144–2150.
Appl. 19, 2–11. Palace, M., Sullivan, F.B., Ducey, M., Herrick, C., 2016. Estimating tropical forest struc­
Liski, J., Kaasalainen, S., Raumonen, P., Akujärvi, A., Krooks, A., Repo, A., Kaasalainen, ture using a terrestrial Lidar. PLoS One 11, e0154115.
M., 2014. Indirect emissions of forest bioenergy: detailed modeling of stump-root Paris, C., Kelbe, D., van Aardt, J., Bruzzone, L., 2015. A precise estimation of the 3D
systems. GCB Bioenergy 6, 777–784. structure of the forest based on the fusion of airborne and terrestrial lidar data. In:
Liu, J., Skidmore, A.K., Wang, T., Zhu, X., Premier, J., Heurich, M., Beudert, B., Jones, S., 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp.
2019. Variation of leaf angle distribution quantified by terrestrial LiDAR in natural 49–52.
European beech forest. ISPRS J. Photogramm. Remote Sens. 148, 208–220. Parker, G.G., Harding, D.J., Berger, M.L., 2004. A portable LIDAR system for rapid de­
Loudermilk, E.L., Hiers, J.K., O’Brien, J.J., Mitchell, R.J., Singhania, A., Fernandez, J.C., termination of forest canopy structure. J. Appl. Ecol. 41, 755–767.
Cropper, W.P., Slatton, K.C., 2009. Ground-based LIDAR: a novel approach to Paynter, I., Saenz, E., Genest, D., Peri, F., Erb, A., Li, Z., Wiggin, K., Muir, J., Raumonen,
quantify fine-scale fuelbed characteristics. Int. J. Wildland Fire 18, 676–685. P., Schaaf, E.S., et al., 2016. Observing ecosystems with lightweight, rapid-scanning
Lovell, J.L., Jupp, D.L.B., Culvenor, D.S., Coops, N.C., 2003. Using airborne and ground- terrestrial lidar scanners. Remote. Sens. Ecol. Conserv. 2, 174–189.
based ranging lidar to measure canopy structure in Australian forests. Can. J. Remote. Paynter, I., Genest, D., Peri, F., Schaaf, C., 2018. Bounding uncertainty in volumetric
Sens. 29, 607–622. geometric models for terrestrial lidar observations of ecosystems. Interface Focus 8,
Luoma, V., Saarinen, N., Kankare, V., Tanhuanpää, T., Kaartinen, H., Kukko, A., 20170043.
Holopainen, M., Hyyppä, J., Vastaranta, M., 2019. Examining changes in stem taper Paynter, I., Schaaf, C., Bowen, J.L., Deegan, L., Peri, F., Cook, B., 2019. Characterizing a
and volume growth with two-date 3D point clouds. For. Trees Livelihoods 10, 382. New England saltmarsh with NASA G-LiHT airborne Lidar. Remote Sens. 11, 509.
Maas, H.G., Bienert, A., Scheller, S., Keane, E., 2008. Automatic forest inventory para­ Pimont, F., Allard, D., Soma, M., Dupuy, J.-L., 2018. Estimators and confidence intervals
meter determination from terrestrial laser scanner data. Int. J. Remote Sens. 29, for plant area density at voxel scale with T-LiDAR. Remote Sens. Environ. 215,
1579–1593. 343–370.
Magney, T.S., Eusden, S.A., Eitel, J.U.H., Logan, B.A., Jiang, J., Vierling, L.A., 2014. Pitkänen, T.P., Raumonen, P., Kangas, A., 2019. Measuring stem diameters with TLS in
Assessing leaf photoprotective mechanisms using terrestrial LiDAR: towards mapping boreal forests by complementary fitting procedure. ISPRS J. Photogramm. Remote
canopy photosynthetic performance in three dimensions. New Phytol. 201, 344–356. Sens. 147, 294–306.
Maguire, A.J., Eitel, J.U.H., Vierling, L.A., Johnson, D.M., Griffin, K.L., Boelman, N.T., Ploton, P., Barbier, N., Momo, S.T., Réjou-Méchain, M., Bosela, F.B., Chuyong, G., Dauby,
Jensen, J.E., Greaves, H.E., Meddens, A.J.H., 2019. Terrestrial lidar scanning reveals G., Droissart, V., Fayolle, A., Goodman, R.C., Henry, M., Kamdem, N.G., Mukirania,
fine-scale linkages between microstructure and photosynthetic functioning of small- J.K., Kenfack, D., Libalah, M., Ngomanda, A., Rossi, V., Sonké, B., Texier, N., Thomas,
stature spruce trees at the forest-tundra ecotone. Agric. For. Meteorol. 269-270, D., Zebaze, D., Couteron, P., Berger, U., Pélissier, R., 2016. Closing a gap in tropical
157–168. forest biomass estimation: taking crown mass variation into account in pantropical
Mäkinen, H., Isomäki, A., 2004. Thinning intensity and long-term changes in increment allometries. Biogeosciences 13, 1571–1585.
and stem form of scots pine trees. For. Ecol. Manag. 203, 21–34. Puttonen, E., Lehtomäki, M., Litkey, P., Näsi, R., Feng, Z., Liang, X., Wittke, S., Pandžić,
Malhi, Y., Jackson, T., Patrick Bentley, L., Lau, A., Shenkin, A., Herold, M., Calders, K., M., Hakala, T., Karjalainen, M., Pfeifer, N., 2019. A clustering framework for mon­
Bartholomeus, H., Disney, M.I., 2018. New perspectives on the ecology of tree itoring circadian rhythm in structural dynamics in plants from terrestrial laser
structure and tree communities through terrestrial laser scanning. Interface Focus 8, scanning time series. Front. Plant Sci. 10, 486.
20170052. Pyörälä, J., Liang, X., Vastaranta, M., Saarinen, N., Kankare, V., Wang, Y., Holopainen,
Marselis, S.M., Tang, H., Armston, J.D., Calders, K., Labrière, N., Dubayah, R., 2018. M., Hyyppä, J., 2018. Quantitative assessment of scots pine (Pinus Sylvestris L.)
Distinguishing vegetation types with airborne waveform lidar data in a tropical whorl structure in a forest environment using terrestrial laser scanning. IEEE J. Sel.
forest-savanna mosaic: a case study in Lopé National Park. Gabon. Remote Sens. Top. Appl. Earth Obs. Remote Sens. 11, 3598–3607.
Environ. 216, 626–634. Pyörälä, J., Kankare, V., Liang, X., Saarinen, N., Rikala, J., Kivinen, V.-P., Sipi, M.,
Martin-Ducup, O., Schneider, R., Fournier, R., 2018. Analyzing the vertical distribution of Holopainen, M., Hyyppä, J., Vastaranta, M., 2019a. Assessing log geometry and wood
crown material in mixed stand composed of two temperate tree species. For. Trees quality in standing timber using terrestrial laser-scanning point clouds. Forestry Int.
Livelihoods 9, 673. J. For. Res. 92, 177–187.
Mengesha, T., Hawkins, M., Nieuwenhuis, M., 2015. Validation of terrestrial laser scan­ Pyörälä, J., Saarinen, N., Kankare, V., Coops, N.C., Liang, X., Wang, Y., Holopainen, M.,
ning data using conventional forest inventory methods. Eur. J. For. Res. 134, Hyyppä, J., Vastaranta, M., 2019b. Variability of wood properties using airborne and
211–222. terrestrial laser scanning. Remote Sens. Environ. 235, 111474.
Metz, J., Seidel, D., Schall, P., Scheffer, D., Schulze, E.-D., Ammer, C., 2013. Crown Qi, C.R., Su, H., Mo, K., Guibas, L.J., 2017a. PointNet: deep learning on point sets for 3D
modeling by terrestrial laser scanning as an approach to assess the effect of above­ classification and segmentation. In: Proceedings of the IEEE Conference on Computer
ground intra- and interspecific competition on tree growth. For. Ecol. Manag. 310, Vision and Pattern Recognition, pp. 652–660.
275–288. Qi, C.R., Yi, L., Su, H., Guibas, L.J., 2017b. PointNet++: Deep hierarchical feature
Miao, Y., Liu, J., Chen, J., Shu, Z., 2020. Structure-preserving shape completion of 3D learning on point sets in a metric space. In: Guyon, I., Luxburg, U.V., Bengio, S.,
point clouds with generative adversarial network. Sci. Sin. Informa. 50, 675–691. Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R. (Eds.), Advances in Neural
Momo Takoudjou, S., Ploton, P., Sonké, B., Hackenberg, J., Griffon, S., de Coligny, F., Information Processing Systems 30. Curran Associates, Inc, pp. 5099–5108.
Kamdem, N.G., Libalah, M., Mofack, G.I.I., Le Moguédec, G., Pélissier, R., Barbier, N., Quegan, S., Le Toan, T., Chave, J., Dall, J., Exbrayat, J.-F., Minh, D.H.T., Lomas, M.,
2018. Using terrestrial laser scanning data to estimate large tropical trees biomass D’Alessandro, M.M., Paillou, P., Papathanassiou, K., Rocca, F., Saatchi, S., Scipal, K.,
and calibrate allometric models: a comparison with traditional destructive approach. Shugart, H., Smallman, T.L., Soja, M.J., Tebaldini, S., Ulander, L., Villard, L.,
Methods Ecol. Evol. 9, 905–916. Williams, M., 2019. The European Space Agency BIOMASS mission: measuring forest
Momo Takoudjou, S., Ploton, P., Martin-Ducup, O., Lehnebach, R., Fortunel, C., Sagang, above-ground biomass from space. Remote Sens. Environ. 227, 44–60.
L.B.T., Boyemba, F., Couteron, P., Fayolle, A., Libalah, M., Loumeto, J., Medjibe, V., Raumonen, P., Kaasalainen, M., Åkerblom, M., Kaasalainen, S., Kaartinen, H., Vastaranta,
Ngomanda, A., Obiang, D., Pélissier, R., Rossi, V., Yongo, O., Sonké, B., Barbier, N., M., Holopainen, M., Disney, M., Lewis, P., 2013. Fast automatic precision tree models
2020. Leveraging signatures of plant functional strategies in wood density profiles of from terrestrial laser scanner data. Remote Sens. 5, 491–520.
African trees to correct mass estimations from terrestrial laser data. Sci. Rep. 10, Raumonen, P., Åkerblom, M., Kaasalainen, M., Casella, E., Calders, K., Murphy, S., 2015.
2001. Massive-scale tree modelling from TLS data. In: ISPRS Annals of Photogrammetry,
Moorthy, I., Miller, J.R., Berni, J.A.J., Zarco-Tejada, P., Hu, B., Chen, J., 2011. Field Remote Sensing & Spatial Information Sciences. 2.
characterization of olive (Olea europaea L.) tree crown architecture using terrestrial Rehush, N., Abegg, M., Waser, L., Brändli, U.-B., 2018. Identifying tree-related micro­
laser scanning data. Agric. For. Meteorol. 151, 204–214. habitats in TLS point clouds using machine learning. Remote Sens. 10, 1735.
Morel, J., Bac, A., Véga, C., 2018. Surface reconstruction of incomplete datasets: a novel Réjou-Méchain, M., Barbier, N., Couteron, P., Ploton, P., Vincent, G., Herold, M., Mermoz,
Poisson surface approach based on CSRBF. Comput. Graph. 74, 44–55. S., Saatchi, S., Chave, J., de Boissieu, F., Féret, J.-B., Takoudjou, S.M., Pélissier, R.,
Morsdorf, F., Schneider, F.D., Gullien, C., Kükenbrink, D., Leiterer, R., Schaepman, M.E., 2019. Upscaling Forest biomass from field to satellite measurements: sources of errors
2020. The Laegeren site: An augmented Forest Laboratory. In: Cavender-Bares, J., and ways to reduce them. Surv. Geophys. 40, 881–911.
Gamon, J.A., Townsend, P.A. (Eds.), Remote Sensing of Plant Biodiversity. Springer Riegler, G., Ulusoy, A.O., Geiger, A., 2017. OctNet: learning deep 3D representations at
International Publishing, Cham, pp. 83–104. high resolutions. In: 2017 IEEE Conference on Computer Vision and Pattern
Murray, J., Fennell, J.T., Blackburn, G.A., Whyatt, J.D., Li, B., 2020. The novel use of Recognition (CVPR), pp. 6620–6629.
proximal photogrammetry and terrestrial LiDAR to quantify the structural complexity Río, M.D., Del Río, M., Bravo-Oviedo, A., Pretzsch, H., Löf, M., Ruiz-Peinado, R., 2017. A
of orchard trees. Precis. Agric. 21, 473–483. review of thinning effects on scots pine stands: from growth and yield to new chal­
Newnham, G., Armston, J., Muir, J., Goodwin, N., Tindall, D., Culvenor, D., Püschel, P., lenges under global change. For. Syst. 26 (eR03S).
Nyström, M., Johansen, K., 2012. Evaluation of Terrestrial Laser Scanners for Rosen, P.A., Hensley, S., Shaffer, S., Veilleux, L., Chakraborty, M., Misra, T., Bhan, R.,
Measuring Vegetation Structure (No. EP124571). CSIRO Sustainable Agriculture Raju Sagi, V., Satish, R., 2015. The NASA-ISRO SAR mission - An international space
Flagship. https://doi.org/10.4225/08/584af4ccc9a19. partnership for science and societal benefit. In: 2015 IEEE Radar Conference
Newnham, G.J., Armston, J.D., Calders, K., Disney, M.I., Lovell, J.L., Schaaf, C.B., (RadarCon), pp. 1610–1613.
Strahler, A.H., Danson, F.M., 2015. Terrestrial laser scanning for plot-scale forest Ross, J., 1981. The Radiation Regime and Architecture of Plant Stands. W. Junk
measurement. Curr. For. Rep. 1, 239–251. Publishers, The Hague.
Nilsson, U., 2010. Thinning of Scots Pine and Norway Spruce Monocultures in Sweden: Saarinen, N., Vastaranta, M., Kankare, V., Tanhuanpää, T., Holopainen, M., Hyyppä, J.,

15
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

Hyyppä, H., 2014. Urban-tree-attribute update using multisource single-tree in­ LiDAR. Remote Sens. Environ. 200, 31–42.
ventory. Forests 5, 1032–1052. Stovall, A.E.L., Anderson-Teixeira, K.J., Shugart, H.H., 2018. Assessing terrestrial laser
Saarinen, N., Kankare, V., Vastaranta, M., Luoma, V., Pyörälä, J., Tanhuanpää, T., Liang, scanning for developing non-destructive biomass allometry. For. Ecol. Manag. 427,
X., Kaartinen, H., Kukko, A., Jaakkola, A., Yu, X., Holopainen, M., Hyyppä, J., 2017. 217–229.
Feasibility of terrestrial laser scanning for collecting stem volume information from Strahler, A.H., Jupp, D.L.B., Woodcock, C.E., Schaaf, C.B., Yao, T., Zhao, F., Yang, X.,
single trees. ISPRS J. Photogramm. Remote Sens. 123, 140–158. Lovell, J., Culvenor, D., Newnham, G., Ni-Miester, W., Boykin-Morris, W., 2008.
Saarinen, N., Kankare, V., Pyörälä, J., Yrttimaa, T., Liang, X., Wulder, M.A., Holopainen, Retrieval of forest structural parameters using a ground-based lidar instrument
M., Hyyppä, J., Vastaranta, M., 2019. Assessing the effects of sample size on para­ (Echidna). Can. J. Remote. Sens. 34, S426–S440.
metrizing a taper curve equation and the resultant stem-volume estimates. Forests 10, Su, H., Maji, S., Kalogerakis, E., Learned-Miller, E., 2015. Multi-view convolutional neural
848. networks for 3D shape recognition. In: 2015 IEEE International Conference on
Saarinen, N., Kankare, V., Yrttimaa, T., Viljanen, N., Honkavaara, E., Holopainen, M., Computer Vision (ICCV), pp. 945–953.
Hyyppä, J., Huuskonen, S., Hynynen, J., Vastaranta, M., 2020. Assessing the effects of Tanhuanpää, T., Yu, X., Luoma, V., Saarinen, N., Raisio, J., Hyyppä, J., Kumpula, T.,
thinning on stem growth allocation of individual scots pine trees. For. Ecol. Manag. Holopainen, M., 2019. Effect of canopy structure on the performance of tree mapping
474, 118344. methods in urban parks. Urban For. Urban Green. 44, 126441.
Sagang, L.B.T., Momo, S.T., Libalah, M.B., Rossi, V., Fonton, N., Mofack, G.I.I., Kamdem, Terryn, L., Calders, K., Disney, M., Origo, N., Malhi, Y., Newnham, G., Raumonen, P.,
N.G., Nguetsop, V.F., Sonké, B., Ploton, P., Barbier, N., 2018. Using volume-weighted Akerblom, M., Verbeeck, H., 2020. Tree species classification using structural fea­
average wood specific gravity of trees reduces bias in aboveground biomass predic­ tures derived from terrestrial laser scanning. ISPRS J. Photogramm. Remote Sens.
tions from forest volume data. For. Ecol. Manag. 424, 519–528. 168, 170–181.
Savage, V.M., Bentley, L.P., Enquist, B.J., Sperry, J.S., Smith, D.D., Reich, P.B., von Tian, J., Dai, T., Li, H., Liao, C., Teng, W., Hu, Q., Ma, W., Xu, Y., 2019. A novel tree
Allmen, E.I., 2010. Hydraulic trade-offs and space filling enable better predictions of height extraction approach for individual trees by combining TLS and UAV image-
vascular structure and function in plants. Proc. Natl. Acad. Sci. U. S. A. 107, based point cloud integration. Forests 10, 537.
22722–22727. Toan, T.L., Le Toan, T., Quegan, S., Davidson, M.W.J., Balzter, H., Paillou, P.,
Schaepman, M.E., Jehle, M., Hueni, A., D’Odorico, P., Damm, A., Weyermann, J., Papathanassiou, K., Plummer, S., Rocca, F., Saatchi, S., Shugart, H., Ulander, L.,
Schneider, F.D., Laurent, V., Popp, C., Seidel, F.C., Lenhard, K., Gege, P., Küchler, C., 2011. The BIOMASS mission: mapping global forest biomass to better understand the
Brazile, J., Kohler, P., De Vos, L., Meuleman, K., Meynart, R., Schläpfer, D., terrestrial carbon cycle. Remote Sens. Environ. 115, 2850–2860.
Kneubühler, M., Itten, K.I., 2015. Advanced radiometry measurements and earth Toshev, A., Szegedy, C., 2014. DeepPose: human pose estimation via deep neural net­
science applications with the airborne prism experiment (APEX). Remote Sens. works. In: 2014 IEEE Conference on Computer Vision and Pattern Recognition, pp.
Environ. 158, 207–219. 1653–1660.
Schneider, F.D., Leiterer, R., Morsdorf, F., Gastellu-Etchegorry, J.-P., Lauret, N., Pfeifer, Trochta, J., Krůček, M., Vrška, T., Král, K., 2017. 3D Forest: An application for descrip­
N., Schaepman, M.E., 2014. Simulating imaging spectrometer data: 3D forest mod­ tions of three-dimensional forest structures using terrestrial LiDAR. PLoS One 12,
eling based on LiDAR and in situ data. Remote Sens. Environ. 152, 235–250. e0176871.
Schneider, F.D., Kükenbrink, D., Schaepman, M.E., Schimel, D.S., Morsdorf, F., 2019. Vaaja, M., Virtanen, J.-P., Kurkela, M., Lehtola, V., Hyyppä, J., Hyyppä, H., et al., 2016.
Quantifying 3D structure and occlusion in dense tropical and temperate forests using The effect of wind on tree stem parameter estimation using terrestrial laser scanning.
close-range LiDAR. Agric. For. Meteorol. 268, 249–257. ISPRS Ann. Photogramm. , Remote Sens. Spatial Inf. Sci 8, 117–122.
Schofield, L.A., Danson, F.M., Entwistle, N.S., Gaulton, R., Hancock, S., 2016. Valbuena, R., O’Connor, B., Zellweger, F., Simonson, W., Vihervaara, P., Maltamo, M.,
Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural Silva, C.A., Almeida, D.R.A., Danks, F., Morsdorf, F., Chirici, G., Lucas, R., Coomes,
networks. Remote Sens. Lett. 7, 299–308. D.A., Coops, N.C., 2020. Standardizing ecosystem morphological traits from 3D in­
Seidel, D., Leuschner, C., Müller, A., Krause, B., 2011. Crown plasticity in mixed forest­ formation sources. Trends Ecol. Evol. 656–667.
s—quantifying asymmetry as a measure of competition using terrestrial laser scan­ Valinger, E., Sjögren, H., Nord, G., Cedergren, J., 2019. Effects on stem growth of scots
ning. For. Ecol. Manag. 261, 2123–2132. pine 33 years after thinning and/or fertilization in northern Sweden. Scand. J. For.
Seidel, D., Fleck, S., Leuschner, C., 2012. Analyzing forest canopies with ground-based Res. 34, 33–38.
laser scanning: a comparison with hemispherical photography. Agric. For. Meteorol. Verbeeck, H., Bauters, M., Jackson, T., Shenkin, A., Disney, M., Calders, K., 2019. Time
154-155, 1–8. for a plant structural economics spectrum. Front. For. Glob. Change 2, 43.
Seidel, D., Schall, P., Gille, M., Ammer, C., 2015. Relationship between tree growth and Vorster, A.G., Evangelista, P.H., Stovall, A.E.L., Ex, S., 2020. Variability and uncertainty
physical dimensions of Fagus sylvatica crowns assessed from terrestrial laser scan­ in forest biomass estimates from the tree to landscape scale: the role of allometric
ning. iForest - Biogeosci. For. 8, 735–742. equations. Carbon Balance Manag. 15, 8.
Seidel, D., Ruzicka, K.J., Puettmann, K., 2016. Canopy gaps affect the shape of Douglas-fir Wallace, L., Lucieer, A., Watson, C., Turner, D., 2012. Development of a UAV-LiDAR
crowns in the western cascades, Oregon. For. Ecol. Manag. 363, 31–38. system with application to forest inventory. Remote Sens. 4, 1519–1543.
Shenkin, A., Chandler, C.J., Boyd, D.S., Jackson, T., Disney, M., Majalap, N., Nilus, R., Wallace, L., Lucieer, A., Watson, C.S., 2014a. Evaluating tree detection and segmentation
Foody, G., Bin Jami, J., Reynolds, G., Wilkes, P., Cutler, M.E.J., van der Heijden, routines on very high resolution UAV LiDAR data. IEEE Trans. Geosci. Remote Sens.
G.M.F., Burslem, D.F.R.P., Coomes, D.A., Bentley, L.P., Malhi, Y., 2019. The world’s 52, 7619–7628.
tallest tropical tree in three dimensions. Front. For. Glob. Change 2, 32. Wallace, L., Musk, R., Lucieer, A., 2014b. An assessment of the repeatability of automatic
Sheppard, J., Morhart, C., Hackenberg, J., Spiecker, H., 2016. Terrestrial laser scanning as forest inventory metrics derived from UAV-borne laser scanning data. IEEE Trans.
a tool for assessing tree growth. iForest-Biogeosci. For. 10, 172. Geosci. Remote Sens. 52, 7160–7169.
Shu, D.W., Park, S.W., Kwon, J., 2019. 3d point cloud generative adversarial network Wallace, L., Gupta, V., Reinke, K., Jones, S., 2016. An assessment of pre- and post fire near
based on tree structured graph convolutions. In: Proceedings of the IEEE surface fuel Hazard in an Australian dry Sclerophyll Forest using point cloud data
International Conference on Computer Vision, pp. 3859–3868. captured using a terrestrial laser scanner. Remote Sens. 8, 679.
Singh, J., Levick, S.R., Guderle, M., Schmullius, C., Trumbore, S.E., 2018. Variability in Wang, D., 2020. Unsupervised semantic and instance segmentation of forest point clouds.
fire-induced change to vegetation physiognomy and biomass in semi-arid savanna. ISPRS J. Photogramm. Remote Sens. 165, 86–97.
Ecosphere 9, e02514. Wang, D., Brunner, J., Ma, Z., Lu, H., Hollaus, M., Pang, Y., Pfeifer, N., 2018. Separating
Singh, J., Levick, S.R., Guderle, M., Schmullius, C., 2020. Moving from plot-based to tree photosynthetic and non-photosynthetic components from point cloud data using
hillslope-scale assessments of savanna vegetation structure with long-range terrestrial dynamic segment merging. For. Trees Livelihoods 9, 252.
laser scanning (LR-TLS). Int. J. Appl. Earth Obs. Geoinf. 90, 102070. Wang, D., Momo Takoudjou, S., Casella, E., 2020a. LeWoS: a universal leaf-wood clas­
Sithole, G., Vosselman, G., 2003. Report: ISPRS Comparison of Filters. Delft University of sification method to facilitate the 3D modelling of large tropical trees using terrestrial
Technology. LiDAR. Methods Ecol. Evol. 11, 376–389.
Smith, D.D., Sperry, J.S., Enquist, B.J., Savage, V.M., McCulloh, K.A., Bentley, L.P., 2014. Wang, D., Schraik, D., Hovi, A., Rautiainen, M., 2020b. Direct estimation of photon re­
Deviation from symmetrically self-similar branching in trees predicts altered hy­ collision probability using terrestrial laser scanning. Remote Sens. Environ. 247,
draulics, mechanics, light interception and metabolic scaling. New Phytol. 201, 111932.
217–229. Wassenberg, M., Chiu, H.-S., Guo, W., Spiecker, H., 2015. Analysis of wood density
Specht, R.L., 1970. Vegetation. In: Leeper, G.W. (Ed.), Australian Environment. profiles of tree stems: incorporating vertical variations to optimize wood sampling
Melbourne University Press, pp. 44–67. strategies for density and biomass estimations. Trees 29, 551–561.
Srinivasan, S., Popescu, S.C., Eriksson, M., Sheridan, R.D., Ku, N.-W., 2014. Multi-tem­ Weber, J., Penn, J., 1995. Creation and rendering of realistic trees. In: SIGGRAPH 1995,
poral terrestrial laser scanning for modeling tree biomass change. For. Ecol. Manag. Proceedings of the 22nd Annual Conference on Computer Graphics and Interactive
318, 304–317. Techniques, pp. 119–128.
Srinivasan, S., Popescu, S., Eriksson, M., Sheridan, R., Ku, N.-W., 2015. Terrestrial laser Wehr, A., Lohr, U., 1999. Airborne laser scanning - an introduction and overview. ISPRS
scanning as an effective tool to retrieve tree level height, crown width, and stem J. Photogramm. Remote Sens. 54, 68–82.
diameter. Remote Sens. 7, 1877–1896. West, G.B., Brown, J.H., Enquist, B.J., 1997. A general model for the origin of allometric
Stobo-Wilson, A.M., Murphy, B.P., Cremona, T., Carthew, S.M., Levick, S.R., 2020. scaling laws in biology. Science 276, 122–126.
Illuminating den-tree selection by an arboreal mammal using terrestrial laser scan­ West, G.B., Enquist, B.J., Brown, J.H., 2009. A general quantitative theory of forest
ning in northern Australia. Remote. Sens. Ecol. Conser. https://doi.org/10.1002/ structure and dynamics. Proc. Natl. Acad. Sci. U. S. A. 106, 7040–7045.
rse2.177. Widlowski, J.-L., Côté, J.-F., Béland, M., 2014. Abstract tree crowns in 3D radiative
Stovall, A.E.L., Shugart, H.H., 2018. Improved biomass calibration and validation with transfer models: impact on simulated open-canopy reflectances. Remote Sens.
terrestrial LiDAR: implications for future LiDAR and SAR missions. IEEE J. Sel. Top. Environ. 142, 155–175.
Appl. Earth Obs. Remote Sens. 11, 3527–3537. Widlowski, J.-L., Mio, C., Disney, M., Adams, J., Andredakis, I., Atzberger, C., Brennan, J.,
Stovall, A.E.L., Vorster, A.G., Anderson, R.S., Evangelista, P.H., Shugart, H.H., 2017. Non- Busetto, L., Chelle, M., Ceccherini, G., Colombo, R., Côté, J.-F., Eenmäe, A., Essery,
destructive aboveground biomass estimation of coniferous trees using terrestrial R., Gastellu-Etchegorry, J.-P., Gobron, N., Grau, E., Haverd, V., Homolová, L., Huang,

16
K. Calders, et al. Remote Sensing of Environment 251 (2020) 112102

H., Hunt, L., Kobayashi, H., Koetz, B., Kuusk, A., Kuusk, J., Lang, M., Lewis, P.E., rainfall interception in urban areas. Landsc. Ecol. Eng. 15, 289–296.
Lovell, J.L., Malenovský, Z., Meroni, M., Morsdorf, F., Mõttus, M., Ni-Meister, W., Yrttimaa, T., Saarinen, N., Kankare, V., Liang, X., Hyyppä, J., Holopainen, M., Vastaranta,
Pinty, B., Rautiainen, M., Schlerf, M., Somers, B., Stuckens, J., Verstraete, M.M., M., 2019a. Investigating the feasibility of multi-scan terrestrial laser scanning to
Yang, W., Zhao, F., Zenone, T., 2015. The fourth phase of the radiative transfer model characterize tree communities in southern boreal forests. Remote Sens. 11, 1423.
intercomparison (RAMI) exercise: actual canopy scenarios and conformity testing. Yrttimaa, T., Saarinen, N., Luoma, V., Tanhuanpää, T., Kankare, V., Liang, X., Hyyppä, J.,
Remote Sens. Environ. 169, 418–437. Holopainen, M., Vastaranta, M., 2019b. Detecting and characterizing downed dead
Wieser, M., Mandlburger, G., Hollaus, M., Otepka, J., Glira, P., Pfeifer, N., 2017. A case wood using terrestrial laser scanning. ISPRS J. Photogramm. Remote Sens. 151,
study of UAS borne laser scanning for measurement of tree stem diameter. Remote 76–90.
Sens. 9, 1154. Yrttimaa, T., Saarinen, N., Kankare, V., Viljanen, N., Hynynen, J., Huuskonen, S.,
Wilkes, P., Lau, A., Disney, M., Calders, K., Burt, A., de Tanago, J.G., Bartholomeus, H., Holopainen, M., Hyyppä, J., Honkavaara, E., Vastaranta, M., 2020. Multisensorial
Brede, B., Herold, M., 2017. Data acquisition considerations for terrestrial laser close-range sensing generates benefits for characterization of managed scots Pine
scanning of forest plots. Remote Sens. Environ. 196, 140–153. (Pinus sylvestris L.) Stands. ISPRS Int. J. Geo-Informa. 9, 309.
Wilkes, P., Disney, M., Vicari, M.B., Calders, K., Burt, A., 2018. Estimating urban above Yu, X., Liang, X., Hyyppä, J., Kankare, V., Vastaranta, M., Holopainen, M., 2013. Stem
ground biomass with multi-scale LiDAR. Carbon Balance Manag. 13, 10. biomass estimation based on stem reconstruction from terrestrial laser scanning point
Wilkes, P., Shenkin, A., Disney, M., Malhi, Y., Boni Vicari, M., 2019. Rapid character­ clouds. Remote Sens. Lett. 4, 344–353.
isation of fine scale branch structure using terrestrial LiDAR. In: Presented at the 4th Yun, T., An, F., Li, W., Sun, Y., Cao, L., Xue, L., 2016. A novel approach for retrieving tree
Scientific Meeting on TLS in Forest Ecology. leaf area from ground-based LiDAR. Remote Sens. 8, 942.
Willim, K., Stiers, M., Annighöfer, P., Ammer, C., Ehbrecht, M., Kabal, M., Stillhard, J., Zhou, Y., Tuzel, O., 2018. VoxelNet: end-to-end learning for point cloud based 3D object
Seidel, D., 2019. Assessing understory complexity in beech-dominated forests (Fagus detection. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern
sylvatica L.) in Central Europe—from managed to primary forests. Sensors 19, 1684. Recognition, pp. 4490–4499.
Wilson, J.W., 1959. Analysis of the spatial distribution of foliage by two-dimensional Zhu, X., Wang, T., Darvishzadeh, R., Skidmore, A.K., Niemann, K.O., 2015. 3D leaf water
point quadrats. New Phytol. 58, 92–99. content mapping using terrestrial laser scanner backscatter intensity with radiometric
Wu, D., Johansen, K., Phinn, S., Robson, A., 2020a. Suitability of airborne and terrestrial correction. ISPRS J. Photogramm. Remote Sens. 110, 14–23.
laser scanning for mapping tree crop structural metrics for improved orchard man­ Zhu, X., Skidmore, A.K., Darvishzadeh, R., Olaf Niemann, K., Liu, J., Shi, Y., Wang, T.,
agement. Remote Sens. 12, 1647. 2018a. Foliar and woody materials discriminated using terrestrial LiDAR in a mixed
Wu, D., Johansen, K., Phinn, S., Robson, A., Tu, Y.-H., 2020b. Inter-comparison of remote natural forest. Int. J. Appl. Earth Obs. Geoinf. 64, 43–50.
sensing platforms for height estimation of mango and avocado tree crowns. Int. J. Zhu, X., Skidmore, A.K., Wang, T., Liu, J., Darvishzadeh, R., Shi, Y., Premier, J., Heurich,
Appl. Earth Obs. Geoinf. 89, 102091. M., 2018b. Improving leaf area index (LAI) estimation by correcting for clumping and
Xi, Z., Hopkinson, C., Chasmer, L., 2018. Filtering stems and branches from terrestrial woody effects using terrestrial laser scanning. Agric. For. Meteorol. 263, 276–286.
laser scanning point clouds using deep 3-D fully convolutional networks. Remote Zianis, D., Muukkonen, P., Mäkipää, R., Mencuccini, M., 2005. Biomass and stem volume
Sens. 10, 1215. equations for tree species in Europe. Silva Fennica Monogr. 4.
Yang, B., Lee, D.K., Heo, H.K., Biging, G., 2019. The effects of tree characteristics on

17