ArticlePDF Available

Abstract and Figures

Tropical forests are currently experiencing large-scale structural changes, including an increase in liana abundance and biomass. Higher liana abundance results in reduced tree growth and increased tree mortality, possibly playing an important role in the global carbon cycle. Despite the large amount of data currently available on lianas, there are not many quantitative studies on the influence of lianas on the vertical structure of the forest. We study the potential of terrestrial laser scanning (TLS) in detecting and quantifying changes in forest structure after liana cutting using a small scale removal experiment in two plots (removal plot and non-manipulated control plot) in a secondary forest in Panama. We assess the structural changes by comparing the vertical plant profiles and Canopy Height Models (CHMs) between pre-cut and post-cut scans in the removal plot. We show that TLS is able to detect the local structural changes in all the vertical strata of the plot caused by liana removal. Our study demonstrates the reproducibility of the TLS derived metrics for the same location confirming the applicability of TLS for continuous monitoring of liana removal plots to study the long-term impacts of lianas on forest structure. We therefore recommend to use TLS when implementing new large scale liana removal experiments, as the impact of lianas on forest structure will determine the aboveground competition for light between trees and lianas, which has important implications for the global carbon cycle.
This content is subject to copyright.
Article
Terrestrial Laser Scanning to Detect Liana Impact on
Forest Structure
Sruthi M. Krishna Moorthy 1,*, Kim Calders 1, Manfredo di Porcia e Brugnera 1,
Stefan A. Schnitzer 2,3 and Hans Verbeeck 1
1CAVElab—Computational and Applied Vegetation Ecology, Ghent University, 9000 Ghent, Belgium;
kim.calders@ugent.be (K.C.); manfredo.diporciaebrugnera@ugent.be (M.d.P.e.B.);
Hans.Verbeeck@UGent.be (H.V.)
2Smithsonian Tropical Research Institute, Balboa Ancon, Apartado 0843-03092, Panama;
Stefan.Schnitzer@mu.edu
3Department of Biological Sciences, Marquette University, Milwaukee, WI 53201-1881, USA
*Correspondence: sruthi.krishnamoorthyparvathi@ugent.be; Tel.: +32-488-97-39-56
Received: 22 March 2018; Accepted: 17 May 2018; Published: 23 May 2018


Abstract:
Tropical forests are currently experiencing large-scale structural changes, including
an increase in liana abundance and biomass. Higher liana abundance results in reduced tree growth
and increased tree mortality, possibly playing an important role in the global carbon cycle. Despite
the large amount of data currently available on lianas, there are not many quantitative studies on
the influence of lianas on the vertical structure of the forest. We study the potential of terrestrial
laser scanning (TLS) in detecting and quantifying changes in forest structure after liana cutting using
a small scale removal experiment in two plots (removal plot and non-manipulated control plot) in
a secondary forest in Panama. We assess the structural changes by comparing the vertical plant
profiles and Canopy Height Models (CHMs) between pre-cut and post-cut scans in the removal plot.
We show that TLS is able to detect the local structural changes in all the vertical strata of the plot
caused by liana removal. Our study demonstrates the reproducibility of the TLS derived metrics for
the same location confirming the applicability of TLS for continuous monitoring of liana removal
plots to study the long-term impacts of lianas on forest structure. We therefore recommend to use
TLS when implementing new large scale liana removal experiments, as the impact of lianas on forest
structure will determine the aboveground competition for light between trees and lianas, which has
important implications for the global carbon cycle.
Keywords:
lianas; TLS; long-term monitoring; tropical forests; aboveground competition; global
carbon cycle
1. Introduction
Aboveground forest structure is an important factor influencing biodiversity, net primary
productivity and the carbon cycle of tropical forests. Tropical forests are undergoing large-scale
structural changes due to anthropogenic disturbances such as increased atmospheric CO
2
, logging,
hunting, and conversion of forested areas into agricultural lands [
1
3
]. One such structural change in
tropical forests is the increase in liana abundance and biomass in the Neotropics. The mechanisms
that explain liana proliferation include increased atmospheric CO
2
, evaporative demands, forest
fragmentation and forest disturbance [
4
]. Lianas are woody climbing plants that use trees and other
plants as structural support for ascending to the canopy. They allocate more resources to canopy
development, stem and root elongation than to their structure, resulting in a high leaf to stem ratio
compared to trees [
5
]. As a result, lianas can significantly attenuate light in the forest and can contribute
up to 40% of the canopy leaf cover [6].
Remote Sens. 2018,10, 810; doi:10.3390/rs10060810 www.mdpi.com/journal/remotesensing
Remote Sens. 2018,10, 810 2 of 19
Due to their erratic growth forms and high abudance, lianas play an important role in tropical
forest ecology and management. For example, lianas with their high canopy to stem ratio can physically
link trees together in the canopy, which serves as a great canopy access system for arboreal animals [
7
,
8
].
Studies have shown that lianas tend to colonize tree fall gaps leading to liana-dominated patches in
the forest [
9
]. Schnitzer and Carson
[10]
showed that gaps maintain better liana diversity and species
richness compared to shade-tolerant tree and overall plant diversity. Both of the above-mentioned
factors play a vital role in logged forest. Logging a tree with inter-crown liana connections could
damage the neighboring trees in the connection and thus result in large canopy gaps [
11
]. In addition,
following logging, lianas in the crown of the logged trees can rapidly regenerate and colonize the gap
and the surrounding trees [12,13].
Pre-felling liana cutting has proven to be a successful management strategy to reduce the impact
of liana-induced damage in a logged forest and to promote tree regeneration in the gaps [
14
17
].
Apart from the forest management strategy, liana-cutting experiments are conducted to study their
long-term impacts on tree-growth and carbon dynamics of the forest [
15
,
18
]. Long-term impact of
liana cutting is often measured by field observations like tree diameter increment, which is then linked
to aboveground biomass using allometric equations [
19
]. Van Der Heijden et al.
[19]
showed that,
after three years of liana removal in large plots in Panama, the net above-ground carbon uptake of
the trees was reduced by 76% in the control plots compared to the removal plots, mainly due to the
liana-induced mortality of trees. Nevertheless, the changes to the vertical and horizontal structure of
the forest, which is directly linked to the aboveground competition between lianas and trees, caused
by liana removal is seldom studied. Pérez-Salicrup
[15]
reported an increase in canopy openness of up
to 4%, measured by hemispherical photographs, 26 months after liana cutting.
Recent advances in remote sensing technologies have enabled us to view the forest structural
complexity in new and unprecedented ways. Terrestrial laser scanning (TLS) is an active remote
sensing technique and can measure various forest structural parameters with high spatial accuracy.
Vertical distribution of Plant Area Index (PAI) and Plant Area Volume Density (PAVD) are key forest
structural metrics directly related to the light interception, growth and primary productivity of the
forest. PAI estimates derived from TLS data were found to be more robust than the PAI estimated from
hemispherical photography [
20
,
21
]. In addition, various studies have successfully derived vertical
profiles of gap fraction, PAI and PAVD for forest ecosystems from TLS data [
22
24
]. Various other
metrics such as the height of maximum vegetation density [
24
], Canopy Height Models (CHM) [
25
],
stem density maps [26] have also been derived from TLS data with high accuracy.
However, only a handful of studies have used TLS data for characterizing the changes in the
forest structure. Calders et al.
[27]
used multi-temporal TLS measurements based on 48 measurement
days from four sampling locations to monitor vegetation phenology by characterizing the shift in
the vertical distribution of PAI. Srinivasan et al.
[28]
have investigated the use of multi-temporal
TLS data to measure the growth of the trees by estimating the change in tree-level aboveground
biomass.
Liang et al. [29]
detected changes to the stems over time caused by natural or other forces
using bi-temporal TLS data. Olivier et al.
[30]
developed a method for quantifying the canopy changes,
mainly the tree crown responses to gap formation by collecting TLS data before and two years after
gap formation. Despite the recent technological advances and different studies highlighting the
impact of lianas on the forest structure, very few studies exist that have tried to analyze the forest
structural change in the 3D canopy structure caused by liana removal. A first study used a Portable
Canopy LiDAR (PCL) [
31
] and Li-COR LAI-2000 plant canopy analyzer (LI-COR, Lincoln, NE, USA)
to characterize the change in forest canopy before and each of the two years after liana cutting [
6
]. PCL
is an upward looking infrared pulsed-laser system that has to be moved through the forest in vertical
transects at a constant pace and spacing. The change in the plant surface density across different
vertical strata between the removal and control plots indicated the contribution of lianas to upper-
and mid-canopy of the forest. However, the contribution of lianas to the upper canopy of the forest
was not quantified owing to the technical limitation of the PCL. Other studies have used VEGNET, a
Remote Sens. 2018,10, 810 3 of 19
phase-based laser rangefinder that reflects the pulses at a fixed angle of 57.5
zenith, to assess whether
liana presence influences the forest successional trajectories by assessing the vertical signature of the
forests of different stand age with and without lianas [32,33].
Since lianas are disturbance-adapted plants, liana abundance is likely to increase with increased
forest disturbance, thereby increasing tree mortality and decreasing tree growth. Our study assesses
the potential of TLS observations of liana removal experiments to better understand the role of lianas
in forest functioning. Forest structural changes owing to liana proliferation could explain aboveground
competition between lianas and trees, which has important implications for the forest carbon cycle [
34
].
In our study, we use for the first time high spatial resolution 3D measurements from a multiple-return
TLS to study lianas in the tropical forest. We collected bi-temporal TLS data from a removal plot before
and six weeks after liana cutting and a control plot of the same size. Based on this test dataset, we
evaluated the potential and limitations of high resolution TLS to observe the vertical and horizontal
structure of lianas in the forest.
The specific objectives of this paper are:
1.
to demonstrate the potential of TLS to detect changes in the vertical structure of the forest before
and after liana removal
2.
to study the reproducibility of the TLS-derived metrics by comparing the structural metrics from
two time steps of the control plot.
2. Materials and Methods
2.1. Study Site
We conducted a liana removal experiment at Gigante Peninsula in Panama (9
9
0
N, 79
51
0
W),
which is adjacent to Barro Colorado Island (BCI) and is a protected forest part of the Barro Colorado
Natural Monument (BCNM). This forest is a 60-year old secondary seasonal tropical forest with
recorded average annual rainfall of 2600 mm per year. The site has a distinct dry period for four
months from December to April during which the rainfall seldom exceeds 100 mm/month [
10
,
19
].
Gigante Peninsula has predominantly clay-rich, highly weathered Oxisols. Tree density and liana
density of the site were found to be 3600 stems and 2000 stems
1 cm per ha in 2008 [
18
]. The map of
the study site with the control and removal plots marked as red dots is shown in Figure 1.
0
N
10°N
90°W
85°W 80°W 75°W
Plot location
500 km
Figure 1.
Map of the study site with the location of the removal and control plots marked with the
red dot. The study site is located in Gigante Peninsula in Panama (9
9
0
N, 79
51
0
W) adjacent to Barro
Colorado Island (BCI).
2.2. Experimental Set-Up
We set up two 10 m
×
10 m plots, one removal plot and one control plot in February 2016.
We measured all trees with diameter
10 cm at breast height (1.3 m from ground) and all lianas with
diameter
1 cm following the liana census protocol in [
35
] before liana removal. The removal and
Remote Sens. 2018,10, 810 4 of 19
control plots were approximately 1.5 km far from each other. Since cutting down lianas could release
the neighboring trees from below-ground competition, we chose the control plot far from the removal
plot [
34
]. The removal plot had a total of five trees and 66 lianas that were rooted in the plot. Eighteen
lianas that were rooted in the plot ascended into the canopy using host trees outside of the 10 m
×
10 m
plot. We recorded the liana load on trees by counting the number of liana stems that rooted within
the plot and attached themselves to the tree trunk or branches in the removal plot. It is possible that
some lianas root outside of the plot and go into the canopy of the trees inside the plot. However, it is
practically very difficult to follow these lianas as they can root far from the host tree. Hence, only the
lianas that root inside the plot are included in the inventory as mentioned in [
35
]. The control plot had
a total of four trees and 45 lianas rooted in the plot. In Table 1, we provide the list of tree diameter,
height and liana load of each tree in the removal plot and the tree diameter and height from the control
plot. We derived the height of the trees using the region-specific diameter at breast height (DBH) to
height allometric equation listed in the study of Van Der Heijden et al.
[19]
. The height derived from
allometric equations do not represent the true height of the trees but rather give an estimate of the
tree height.
Table 1.
Overview of trees, diameter and number of liana stems on each trees in the removal and
control plot. Trees 2 and 3 are the same individual that branched below 1 m from the ground.
Plot Type Tree No. Diameter at
Breast Height (cm) Height (m) No. of Liana Stems
on the Tree
Basal Area of
Liana Load (cm2)
Removal
Plot
1 10.9 12.3 5 9.09
2 16.1 15.3 18 60.83
3 12.6 13.3
4 43.7 25.2 10 83.13
5 15.3 14.9 18 96.29
Control
Plot
1 47.4 26.1 - -
2 11.7 12.8 - -
3 21.2 17.7 - -
4 15.5 14.9 - -
Trees 2 and 3 were actually the same individual that branched below 1 m from the ground.
As a result
, we counted the lianas load on one tree also for the other, as it was difficult to distinguish
the canopy of these two trees from the ground. Lianas were cut near the forest floor using a machete
and most of the liana stems in the understorey were moved without damaging the tree crowns on
23 February 2016
as it can be seen in Figure 2. Lianas in the area of 2 m around the plot were also cut
to avoid the edge effects.
Figure 2. Comparison of the removal plot before (left) and after (right) liana removal.
Remote Sens. 2018,10, 810 5 of 19
2.3. LiDAR Data
We collected TLS data from the two 10 m
×
10 m plots, using four scan locations at the corners of
the plot. TLS data was collected twice in the removal plot, once on 22 February 2016 one day before
liana removal and again on 3 April 2016, six weeks after liana removal, assuming that liana leaves
would have mostly fallen off the canopy in the six week time period [
6
]. We scanned the control plot
of size 10 m
×
10 m twice with the same time gap of six weeks (26 February 2016–4 April 2016) to
account for natural leaf abscission.
We used a RIEGL VZ400 terrestrial laser scanner (RIEGL Laser Measurement Systems GmbH,
Horn, Austria), which is a multiple return time-of-flight based scanner using a narrow infrared laser
beam of wavelength 1550 nm and a beam divergence of 0.35 mrad. We mounted the scanner on a
tripod at approximately 1.3 m from the ground and used an angular sampling resolution of 0.02
.
The acquisition time for one full scan with an angular resolution of 0.02
is 12 min 19.6 s. Figure 3
illustrates the view from one scanning position from the removal plot before and after liana removal in
a cylindrical projection. Figure 3b of the post-cut TLS data clearly shows the missing liana tangles after
liana cutting. Similarly Figure 4illustrates view from one scanning position for the control plot. As
Figure 4b indicates, the TLS data collected on 4 April 2016 indicates displacement of some vegetation
such as palm leaves due to wind or arboreal animals.
(a)Before removal (b)After removal
Figure 3.
View of terrestrial LiDAR data from one scan position of the removal plot with azimuth
ranging from 120
to 240
and zenith ranging from 30
to 130
. The data is colored according to the
range of the laser pulse.
(a)Terrestrial LiDAR data on 26 Feb. 2016 (b)Terrestrial LiDAR data on 4 Apr. 2016
Figure 4.
View of terrestrial LiDAR data from one scan position of the control plot with azimuth
ranging from 0
to 100
and zenith ranging from 30
to 130
. The data is colored according to the range
of the laser pulse.
Remote Sens. 2018,10, 810 6 of 19
As the scanner has a zenith range of only 100
from 30
to 130
, to get a full hemispherical view of
the canopy, the scanner is tilted to 90
from the vertical axis for an additional tilted scan. We distributed
reflective targets in the field to register the upright scan with the corresponding tilt scan and also to
co-register all four scan locations. The registration of all four positions for the removal and control
plots was done using the RISCAN Pro software (version 2.5.3, RIEGL Laser Measurement Systems
GmbH, Horn, Austria) provided by Riegl. After registration, the raw TLS data were exported as ASCII
files for deriving CHMs and stem maps. In addition, the raw TLS data were converted into an open
source Sorted Pulse Data (SPD) format and analysed using the open source LiDAR data processing
library PyLidar for deriving the vertical profiles of Pgap, PAI and PAVD [36,37].
2.4. Co-Registration of Bi-Temporal Data
We co-registered the TLS data before and after liana removal in the removal plot and two scans
from the control plot to facilitate the comparison of the all TLS-derived metrics from two different
time steps. The point cloud of the region of interest from the two scans of removal and control plot
were exported as an ASCII file from RiSCAN pro. The co-registration was done on these ASCII
files using an Iterative Closest Point (ICP) algorithm implemented in CloudCompare (version 2.8.1,
CloudCompare, GPL software) [
38
]. The detailed steps of co-registration of the point clouds are
described in Appendix A.2. We refer to the normalized
z
-value as height in this article. Since the
topography of the plots studied are relatively flat, we went for a simple plane-fitting normalization as
explained in [24].
2.5. Vertical Plant Profiles
The frequency distribution of all the returns with respect to height for the bi-temporal data of
the removal and the control plot were estimated using a Gaussian kernel density estimator with
a bandwidth of 0.5. We also estimated the frequency distribution of the different returns (first, second,
third and fourth) with respect to height for the pre- and post-cut data of the removal plot using the
kernel density estimator with the same bandwidth of 0.5.
We performed a two-sided Kolmogorov–Smirnov test (KS test) to test the equality of the relative
frequency distribution of height of the two time steps for the removal as well as control plot [
39
].
We chose this test as it does not assume that the data are sampled from a defined distribution like
Gaussian distribution.
We described the vertical plant profiles using the vertically resolved gap probability, PAI and
PAVD derived from the TLS data. As mentioned in the Section 2.3, we used the open source
python library PyLidar for deriving the vertical profiles. We followed the method described in
Calders et al. [24]
to estimate the Pgap, PAI and PAVD profiles. PAI(z), PAI as a function of height z,
is calculated through the estimates of Pgap at the corresponding height. PAVD(z) is calculated based
on vertically averaged PAI at 57.5
zenith and Pgap(z). The basic steps and formulas are described
in Appendix A.1. For the removal and control plot, we derived the profiles for the specific azimuth
region of one scan location shown in Figures 3and 4, respectively, that fell within the 10 m
×
10 m plot.
We did not derive the profiles for the data from other scan locations as the scans had many objects
too close to the scanner (<1.5 m from the scanner), especially in pre-cut scan, leading to artificial gaps
resulting in underestimated PAI.
2.6. Nearest Neighbor Distance
The angular resolution (0.02
in this case) and the range, the distance between the target and
the instrument, determine the vertically resolved Average Nearest Neighbor (ANN) distance for
the point cloud. The minimum of the average nearest neighbor distance at a particular range is the
smallest distance resolvable by the TLS instrument (d
min
), which depends on instrument-specific beam
divergence [40]. Vertically resolved ANN is calculated as in Equation (1):
Remote Sens. 2018,10, 810 7 of 19
ANN(z) =1
N(z)
N(z)
i=1
NN(pi),{pi: z <zpi<z+dz}. (1)
NN is the nearest neighbor distance for a point in the point cloud. The distance between two
points in the point cloud for a plot is defined by the Euclidean distance between the two points in 3D.
N(z) is the total number of points in the height interval between z and z + dz. For every 1 m z-bin,
we calculate the average distance to four nearest neighbors for all the points falling in the bin.
The ANN(z) values indicate the average distance between the topologically connected points
from forest floor to canopy. In forests with dense vegetation, ANN(z) is not only influenced by d
min
,
but also by the spatial distribution of the forest elements occluding the other elements with increasing
height [
40
]. In our study, we compare the ANN(z) between the pre- and post-liana cut TLS data from
the removal plot to assess the complex spatial distribution of liana stems and leaves in the plot. We also
compare the ANN(z) between the multiple return and the first return TLS data to assess the advantage
of using multiple-return scanners over the first-return scanners for studying lianas in the tropics.
2.7. Canopy Height Models
Canopy Height Models describe the horizontal distribution of the height of a forest canopy as a
three-dimensional surface [
41
]. In this study, we generated CHMs to compare the change in height
in the observed canopy of the co-registered point clouds in the removal plot before and after liana
removal. We also generated CHMs for the two point clouds from the control plot and compare it to
that of the removal plot. Bi-temporal TLS data from the removal and control plots were co-registered
as described in Appendix A.2. We derived CHMs by selecting the points of highest
z
-value within each
50 cm
×
50 cm x,y grid. After generating CHMs for the two point clouds, we calculated the absolute
difference between the corresponding grids of the bi-temporal CHMs to compare the change in the
spatial distribution of the observed canopy height for both the removal plot and control plot.
3. Results and Discussion
3.1. Vertical Plant Profiles
The frequency distribution of returns with respect to height shows major differences in the
number of returns from understorey and the upper canopy of the bi-temporal scans in the removal
plot (Figure 5a). After cutting lianas, we found a 10% decrease in the absolute number of returns
from the first five meters and a 56% increase in the absolute number of returns above 15 m post liana
cutting. The mid-canopy shows very little change before and after liana cutting in terms of the vertical
distribution of returns. This is likely because the liana stems in the mid-canopy remained, whereas the
liana stems in the understorey were removed while cutting and the liana leaves in the upper canopy
had mostly fallen six weeks after liana cutting. In the pre-cut data, there were fewer returns from the
upper canopy with lower canopy height compared to the post-cut data because of the occlusion of
liana leaves obstructing the pulses emitted.
The pre- and post-cut frequency distributions of height were significantly different in the removal
plot (KS statistic = 0.16, P< 0.001), but not in the control plot as it can be seen from Figure 5b (KS
statistic = 0.05, P< 0.144).
We found differences in the frequency distribution of height of the different returns in the
understorey and the canopy between the pre- and post-cut TLS data (Figure 6). For instance, after liana
cutting, the absolute number of returns decreased by 60% in the understorey up to five meters and
increased by 70% above 15 m in the post-cut TLS data when only the last returns (fourth return) were
considered. Results of the frequency distribution of heights of different returns for the bi-temporal
data of the control plot show similar results as in Figure 5b and are therefore not discussed further.
However, the results are included in Appendix A.3 for reference.
Remote Sens. 2018,10, 810 8 of 19
0 5 10 15 20 25
Height (m)
0.00
0.02
0.04
0.06
0.08
0.10
0.12
Relative frequency
Pre-cut (Feb 22, 2016)
Post-cut (Apr 3, 2016)
(a)Removal plot
0 5 10 15 20 25
Height (m)
0.00
0.02
0.04
0.06
0.08
0.10
0.12
Relative frequency
Feb 26, 2016
Apr 4, 2016
(b)Control plot
Figure 5.
Comparison of frequency distribution with Gaussian kernel density estimation of all vertical
returns combined for the bi-temporal TLS data from the removal (a) and control plot (b).
0 5 10 15 20
Height (m)
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
Relative frequency
First return
Second return
Third return
Fourth return
(a)Before removal
0 5 10 15 20
Height (m)
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
Relative frequency
First return
Second return
Third return
Fourth return
(b)After removal
Figure 6.
Comparison of the frequency distribution with Gaussian kernel density estimation of different
vertical returns with height between the pre-cut (a) and post-cut (b) data of the removal plot.
We derived the vertical profiles of PAVD with both first-return and multiple-return TLS data for
the pre- and post-cut scans of the removal plot (Figure 7). The profiles were generated for the azimuth
region shown in Figure 3. The total PAI decreased from 6.5 to 6.3 after liana cutting. We found 48%
decrease in the first five meters, 18% decrease at 15 m and 90% increase above 16 m in PAVD post-liana
cutting when using multiple returns. Both the first return and multiple return PAVD profiles show
very little change in the mid-canopy before and after liana cutting similar to the vertical distribution of
returns shown in Figure 5a. Both the profiles show increased PAVD beyond 16 m. This increase in
PAVD beyond 16 m mainly corresponds to the occlusion of liana leaves in the canopy at about 15 m.
Our results demonstrate that liana significantly increases occlusion in the upper canopy owing to their
high leaf to stem ratio [5].
Similar to Calders et al.
[24]
, we observed a larger number of laser pulse returns at greater canopy
height when using multiple returns as shown in Figures 6and 7. This highlights the advantage of using
a multiple return instrument in forest ecosystems with dense understory and high liana abundance.
We recorded 270% more returns from the canopy (
15 m) when using multiple returns compared
to first returns in the pre-cut and post-cut TLS data. In addition, the highest return of pulses for
the pre-cut data went down to 18 m from 20 m when only first returns were considered. For the
post-cut data, the highest return went down to 19.5 from 22 m. A two meter increase in TLS estimated
height post liana cutting clearly indicates that the complex, spatial distribution of different forest
Remote Sens. 2018,10, 810 9 of 19
elements results in increased occlusion effects with increased height. The height of the largest tree
according to region-specific diameter to height allometric equation was 25.19 m and 26.11 m in the
removal and control plot, respectively. The maximum height reached by the multiple return TLS used
in the study was 22 m (after liana cutting) and 24 m in the removal and control plot, respectively.
There are two plausible explanations for this height difference of 3 m in the removal plot and 2 m in the
control plot. One possible explanation is that the TLS is underestimating the maximum canopy height
owing to tree top occlusions. Tree height measurements from vertex rangefinders have proved to be
more accurate in some forest ecosystems compared to TLS and other traditional height measurement
methods [
42
44
]. However, this is not always the case as TLS has been proved to reach the top of the
canopy easily in other forest ecosystems, where the TLS-derived tree height was compared to true tree
height from destructive measurements [
45
]. The other explanation is that the allometric equation used
is overestimating the tree height.
0.0 0.5 1.0
PAVD
(
m
2/
m
3)
0
5
10
15
20
25
Height (m)
First return
Pre-cut
Post-cut
0.0 0.5 1.0
PAVD
(
m
2/
m
3)
0
5
10
15
20
25 Multiple return
Pre-cut
Post-cut
Removal plot - PAVD Profile
Figure 7.
Comparison of the first-return and multiple-return Plant Area Volume Density (PAVD)
profiles of the removal plot.
The PAVD profile with only first returns demonstrates more vegetation closer to the scanner in the
understory than in the upper canopy compared to the multiple-return PAVD profile. The peak PAVD
height before liana cutting remains unchanged at 15.5 m for both first and multiple return profiles.
This demonstrates that both the first return and multiple return profiles agree on the height of higher
concentration of liana leaves. However, the peak PAVD height after liana cutting went down to 13.5 m
from 15 m demonstrating the downside of first return TLS instruments of having a low amount of
returns from the canopy despite the mitigated occlusion effects from the loss of other canopy elements.
The bi-temporal vertical profiles of PAVD of the control plot are shown in Figure 8a. Figure 8b
shows the box plot of absolute difference between the PAVD of two time steps of the control and
removal plot. As it can be seen from Figure 8b, the bi-temporal PAVD profiles of the control plot did not
show any major shifts. Our results demonstrate not only the reproducibility of the TLS-derived metrics
to do repeated measurements to study the temporal dynamics of tropical forest structure as a function
of height, but also the potential of TLS to determine the changes in the vertical profile followed by liana
removal. Reproducibility of TLS-derived metrics has also been confirmed in a broadleaf deciduous
forest by a spring phenology monitoring study of Calders et al. [27].
We only used one location to derive the vertical profiles of PAVD in the removal plot and control
plot. This was mainly because of many objects being close to the scanner (<1.5 m from scanner)
especially in the pre-cut scan indicated by the blue zones in Figure 9. Including the blue zones in
the zenith band of 55–60
for the estimation of Pgap will result in an overestimated Pgap and thus
an underestimated PAI. One way to overcome this is to scan from a location with no interfering
Remote Sens. 2018,10, 810 10 of 19
vegetation too close to the scanner or to enable the near range activation mode (which is scanner
specific,
0.5 m for
our instrument). The other way is to correct for these effects using voxel-based
methods [46,47], which will be an area of future study.
0.0 0.5 1.0
PAVD
(
m
2/
m
3)
0
5
10
15
20
25
Height (m)
First return
Feb 26, 2016
Apr 4, 2016
0.0 0.5 1.0
PAVD
(
m
2/
m
3)
0
5
10
15
20
25 Multiple return
Feb 26, 2016
Apr 4, 2016
Control plot - PAVD Profile
(a)Control plot PAVD profile
Control plot Removal plot
0.00
0.05
0.10
0.15
0.20
0.25
0.30
Absolute PAVD difference (
PAVDt
2
PAVDt
1 )
(b)Absolute PAVD difference: removal vs control plot
Figure 8.
(
a
) shows the bi-temporal Plant Area Volume Density (PAVD) profiles of the control plot and
(
b
) shows the absolute PAVD difference box plot of the control and removal plot, with t
2
being the data
from second time step and t1, the first time step.
Figure 9.
TLS data before liana removal with azimuth ranging from 0
to 360
and zenith ranging from
30
to 130
. The data is colored according to range of the laser pulse with black being low and white
being high. The areas with blue indicates regions of no returns either because the object was too close
to the scanner (<1.5 m) or because of a gap.
3.2. Nearest Neighbor Distance
Comparison of the multiple-return and the first return ANN (z) for the pre- and post-cut TLS
data of the removal plot and the bi-temporal TLS data of the control plot are shown in Figure 10.
Lower ANN value for a particular height compared to the corresponding d
min
indicates theoretical
oversampling and higher ANN value indicates theoretical undersampling. The line representing the
laser beam diameter at different heights shows what theoretically can be resolved by the laser beam.
The higher the deviation from this line, the higher is the occlusion and lower is the quality of the data.
Pre-cut TLS data clearly shows higher effects of occlusion than the post-cut TLS data in the
upper canopy, especially after 15 m. These effects are intensified when using only first returns as
the ANN increases by 2 cm from 1 cm and 6 mm from 4 mm between 15 m and 17 m in the pre-cut
data and post-cut data, respectively. It is evident from PAVD profiles of Figure 7that the highest
concentration of liana leaves is located at about 15 m leading to an increased ANN above 15 m in the
pre-cut data. This implies that, in the first return pre-cut data, the size of the smallest object that could
be resolved at 17 m is at least 3 cm and there are not enough topologically connected objects beyond
17 m, whereas with multiple returns the minimum size of the object resolvable at 18 m is 2.5 cm. As this
Remote Sens. 2018,10, 810 11 of 19
trend is only expected to increase when we move higher up in the canopy, multiple-return TLS has
an advantage over first-return TLS to study impact of lianas on forest structure as liana leaves are
mainly in the canopy.
Bi-temporal TLS data from the control plot shows no significant difference in the vertical ANN
profile. The multiple return ANN at 21 m remains at 4.1 cm and 4.2 cm for the TLS data collected on
26 February 2016
and 4 April 2016, respectively. The same is true for the first return ANN at 19 m,
which stays at 3 cm and 2.9 cm for the TLS data on
26 February 2016
and 4 April 2016, respectively.
This again demonstrates the potential of TLS to produce reproducible results and to determine the
changes in the vertical profile of the forest. However, this also shows the occlusion effects of dense
vegetation beyond 15 m, which should be addressed in the upcoming studies focusing on lianas in the
tropical biomes.
0.00 0.01 0.02 0.03
ANN (m)
0
5
10
15
20
z (m)
Feb 22, 2016
0.00 0.01 0.02 0.03
ANN (m)
0
5
10
15
20 Apr 3, 2016
(a)Removal plot
0.00 0.02 0.04
ANN (m)
0
5
10
15
20
z (m)
Feb 26, 2016
0.00 0.02 0.04
ANN (m)
0
5
10
15
20
Apr 4, 2016
(b)Control plot
Figure 10.
Comparison of multiple return vs. first return vertically resolved average nearest neighbor
distance (ANN(z)) for pre- and post-cut TLS data of the removal plot (
a
) and bi-temporal data of the
control plot (b). (Laser beam diameter, ?First return and +Multiple return).
3.3. Canopy Height Models
The difference between the pre- and post-cut CHMs of the removal plot and the bi-temporal scans’
CHMs of the control plot are illustrated in Figure 11. We observed an overall increase in canopy height
in the CHM after liana cutting with 82% of the grids seeing an increase in height between 0 to 6 m and
18% of the grids losing vegetation between 0 and 4 m. There were some local changes in the CHM
between two time steps in the control plot. The observed increase in canopy height was approximately
the same as the observed decrease in the canopy height in the control plot with 54% of the grids of
the TLS data from 4 April 2016 gaining height between 0 and 4 m and 46% of the grids losing height
between 0 and 4 m. The average observed increase in canopy height was 1.5 m in the removal plot and
0.4 m in the control plot, whereas the average observed height loss was 0.5 m for both the removal and
control plot.
It can be seen from Figure 11 that contiguous grids with observed increase in canopy height
ranging between 1 and 6 m are co-located with the location of trees 4 and 5 in the removal plot.
We interpreted the virtual height gain in these contiguous grids as loss of liana leaves in those grids.
Clark and Clark
[48]
estimated the liana load on trees by two measures: the number of liana stems
attached to the tree trunk and Crown Occupancy Index (COI). The COI is estimated on a scale of
five based on the percentage of the tree crown occupied by liana leaves (0, 1–25%, 26–50%, 51–75%,
76–100%). They showed that the COI of trees were positively correlated with the basal area of liana
load attached to their trunks. We found evidence for their hypothesis as trees 4 and 5 had higher basal
area of liana load than the other trees in the plot corresponding to loss of more liana leaves in the
canopy. Figure 12 shows the height at which these changes occurred pre- and post-liana cutting in
the removal plot. We showed that most of the observed height gain of more than 2 m occur beyond
Remote Sens. 2018,10, 810 12 of 19
14 m in the pre-cut data, which is also evident from PAVD profiles in Figure 7. As tree 5 is a small
tree with 15.3 cm DBH and 14.9 m height, it is reasonable to assume that most of the height gain
occurred within the crown of tree 4, the largest tree of the plot. Our results are in agreement with the
other studies, which showed lianas compete more with upper canopy trees for light by colonizing the
tree crowns [49,50] and the study of Clark and Clark [48] that showed the liana loads to be positively
correlated with tree diameter.
Difference between CHMs of the control plot showed local changes in the canopy. These changes
might be either due to the displacement of vegetation or due to loss of some leaves or branches within
the six-week period as indicated by Figure 4. However, as Figure 11b,d indicate, the scatterplots
confirm the height agreement of bi-temporal data of the control plot as opposed to the bi-temporal
data of the removal plot. The TLS instrument used in the study shows the potential for capturing the
spatial distribution of changes in the canopy in high detail, which has important implications for the
long-term monitoring of liana removal experiments with both removal and control plots [6].
1086420
x (m)
0
2
4
6
8
10
y (m)
1
2
3
4
5
CHMApr
3, 2016
CHMFeb
22, 2016
<
-4
-3
-2
-1
0
1
2
3
> 4
(a)Removal plot
10 12 14 16 18 20
Maximum height (m) - Feb 22, 2016
10
12
14
16
18
20
22
Maximum height (m) - Apr 3, 2016
1:1 line
y (m) = 0.96 * x (m) + 1.38
(b)Differences in height (m)—before vs after removal
0 2 4 6 8 10
x (m)
10
8
6
4
2
0
y (m)
12
3
4
CHMApr
4, 2016
CHMFeb
26, 2016
<
-4
-3
-2
-1
0
1
2
3
> 4
(c)Control plot
10 12 14 16 18 20 22 24
Maximum height (m) - Feb 26, 2016
10
12
14
16
18
20
22
24
Maximum height (m) - Apr 4, 2016
1:1 line
y (m) = 0.97 * x (m) + 0.55
(d)Differences in height (m)—26 Feb. 2016 vs. 4 Apr. 2016
Figure 11.
Difference map of the Canopy Height Models (CHM) with circles representing the location
of trees
10 cm in DBH. (
a
) removal plot and (
c
) control plot. The size of the circles correspond to
the size of the trees. The color scale represents the change in height in meters between the two time
steps. Scatter plot showing the CHM difference between bi-temporal TLS data for removal plot (
b
) and
control plot (d).
Remote Sens. 2018,10, 810 13 of 19
10 12 14 16 18 20 22
Pre-cut observed canopy height (m)
10
12
14
16
18
20
22
Post-cut observed canopy height (m)
1:1
Figure 12.
Differences in Canopy Height Models (CHM) for CHM
post-cut
CHM
pre-cut
2 m in the
removal plot.
3.4. Limitations of the Study
As explained in the previous sections, occlusion is a challenging problem when deploying TLS
in complex environments like tropical forests. The observed increase in canopy height after liana
cutting in Figure 11a does not correspond to the real increase in canopy height but rather indicates
the occlusion effects of liana leaves in the pre-liana cut CHM. We collected TLS data only from the
four corners of the plot. One of the ways to overcome the occlusion is to have a better sampling
strategy by increasing the number of scanning locations. Increasing the number of scans around the
plot enables us to cover the plot from multiple viewing angles. In addition, scanning the plot from
longer distances (approximately 5–10 m from the plot edge) could result in getting better views of
the canopy [
51
]. Our study was able to detect the changes in the vertical structure of the forest, but
a better sampling strategy could result in data with better quality enabling the quantification of the
changes with higher accuracy. In addition, future development of TLS instruments with smaller beam
divergence could help in resolving smaller objects at greater heights in the canopy, thereby mitigating
the occlusion effects.
Another important limitation of the study is the plot size and sample size. We used a small sample
size because our study is a first exploratory effort to test the potential of TLS to study the effect of
lianas on forest structure. We tried to balance the number of lianas that we sacrificed for this study
by cutting enough lianas to rigorously see the effect of liana cutting on forest structure, but not so
many that it would have a huge impact on natural forest regeneration and functioning. Though our
study cannot conclude about the ecological impact of lianas on forest structure, the changes observed
in the PAVD profiles, ANN(z) profiles and the CHMs of the removal plot derived from the TLS data
have important implications for the future studies of liana impact on the forest structure. Collecting
high quality TLS data for large areas in dense forest ecosystems can be time-consuming. For instance,
scanning a 1 ha plot with scan positions located every 10 m and few scan positions located outside
of the plot boundaries in a dense tropical forest could take up to five days including the time taken
for preparing the plot (distributing the control reflective targets for registration). However, such high
detailed data can yield answers to some of the unsolved ecological questions about liana influence on
forest structure.
Remote Sens. 2018,10, 810 14 of 19
4. Conclusions
As lianas strongly compete with trees for light, studying the changes in the vertical forest structure
with and without lianas over the long term could explain how lianas impact the forest structure.
We demonstrate the potential of TLS to study the changes in all the vertical strata of the forest caused
by liana removal. Our results demonstrate the advantages of having multiple returns over first
returns to quantify changes in the upper canopy of the forest, especially in environments where the
understorey is dense or lianas are abundant. Our results also highlight the reproducibility of TLS
derived metrics to do repeated long-term measurements of the liana removal and control plots. Though
collecting TLS data from large areas in complex environments can be time-consuming, lessons learned
from the current small scale study will enable us to collect high quality data in future. We highly
recommend to include TLS observations in future large scale liana removal experiments in the long
term to quantitatively study the impact of lianas on the forest structure.
Author Contributions:
S.M.K.M., M.d.P.e.B., S.S. and H.V. conceived and designed the experiments; S.M.K.M.,
M.d.P.e.B. and S.S. performed the experiments; S.M.K.M. and K.C. analyzed the data; S.M.K.M., K.C., and H.V.
wrote the paper. All authors discussed the results and commented on the manuscript.
Acknowledgments:
The research leading to these results was funded by the European Research Council Starting
Grant 637643 (TREECLIMBERS) and National Science Foundation grants: NSF-IOS 1558093 and NSF-DEB 1822473.
We thank Iebe Stroobants, Elizabeth Kearsley, Maria M. García León, Felipe Nery Arantes Mello, Claudio Manuel
Monteza and Patrick Cvecko for their assistance with fieldwork. The authors would like to thank the Smithsonian
Tropical Research Institute (STRI) for their support and for allowing us to perform liana removal experiments
in Gigante.
Conflicts of Interest: The authors declare no conflict of interest.
Abbreviations
The following abbreviations are used in this manuscript:
TLS Terrestrial Laser Scanning
DBH Diameter at Breast Height
PAI Plant Area Index
LAI Leaf Area Index
PAVD Plant Area Volume Density
CHM Canopy Height Model
LiDAR Light Detection and Ranging
PCL Portable Canopy LiDAR
COI Crown Occupancy Index
BCI Barro Colorado Island
BCNM Barro Colorado Natural Monunment
SPD Sorted Pulse Data
ICP Iterative Closest Point
KS Kolmogorov–Smirnov
ANN Average Nearest Neighbor
Appendix A
Appendix A.1. Vertical Profiles of Pgap, PAI and PAVD
The basic steps and formulas for deriving the vertical profiles of Pgap, PAI and PAVD based on
Calders et al. [24] are described in this section.
Pgap (z) from a multi-return LiDAR is estimated using Equation (A1):
Pgap(¯
θ, z) = 1wi(ziz, ¯
θ)
N(¯
θ), w =1
ns. (A1)
Remote Sens. 2018,10, 810 15 of 19
z is the height above ground.
¯
θ
is the mid-point of the finite zenith angle interval used to aggregate
the gap probability at a certain height z. N(
¯
θ
) is the total number of laser pulses sent for the zenith
angle interval and
ns
is the total number of returns for that transmitted pulse. In this study, we used
the zenith angle between 5
and 60
and a zenith angle interval of 5 degrees. We only used a 90
azimuth for every scan position that fell within the 10 m ×10 m plot area.
PAI (z) and total PAI are calculated as in Equations (A2) and (A3), respectively:
PAI(z) = 1.1 log Pgap(57.5, z), (A2)
PAI =PAI(H). (A3)
H is the canopy height. PAI is calculated at the hinge region (57.5
), by using the laser pulses
transmitted and reflected within the zenith band of 55
and 60
. The hinge region is used mainly
because the foliage distribution function is assumed to be invariant in this region.
PAVD(z) is calculated using the PAI at the hinge region and the Pgap(z) based on Equation (A4)
described in Jupp et al. [23]:
PAVD(z) = PAI δ
δz
log Pgap(¯
θ, z)
log Pgap(¯
θ, H). (A4)
In addition to restricting the pulses transmitted and reflected within a specific azimuth falling
within the area of the plot, we also excluded those azimuth angle intervals from the scans that had no
returns registered because an object was too close to the scanner (within 1.5 m from the scanner).
Vertical profiles of Pgap derived for the hinge region between 55
and 60
for the pre-cut
and post-cut scans of the removal plot and the bi-temporal scans of the control plot are shown
in Figure A1a,b, respectively.
(a)Removal plot
0.0 0.2 0.4 0.6 0.8 1.0
Pgap
(57.5
o
)
0
5
10
15
20
25
30
Canopy
height (m)
Control plot -
Pgap
(57.5
o
) profile
Feb 26, 2016
Apr 4, 2016
(b)Control plot
Figure A1. Pgap profile in the hinge region (57.5zenith).
Appendix A.2. Co-Registration of the Bi-Temporal Scans
Co-registration was done as follows:
1.
We isolated the ground points from the TLS data of two time steps for both the removal and
control plot using the Cloth Simulation Filter (CSF) algorithm implemented in CloudCompare.
CSF is a tool to extract ground points from a discrete return LiDAR data [52].
2.
We derived stem maps from TLS for the two time steps following the method mentioned in
Section 2.7.
3.
We used the stem points plus the ground points from these two pieces of different temporal
data as input for the first coarse manual registration in CloudCompare with one point cloud as
reference and the other as the one to be aligned.
Remote Sens. 2018,10, 810 16 of 19
4.
We then applied the transformation matrix from the manual registration to the whole point cloud
to be aligned.
5.
We used the ICP algorithm [
53
] implemented in CloudCompare for fine registration after the
first coarse manual registration. We selected all the points from the ground up to 4 m for
fine registration.
6.
We applied the transformation matrix that resulted from the ICP fine registration to the whole
point cloud to be aligned.
Co-registration of the bi-temporal TLS data was achieved with an RMSE of 0.08 m in the removal
plot and 0.06 m in the control plot, respectively. Figure A2 shows the result of co-registration using
a tree trunk from the removal plot.
(
a
)Pre-registration (
b
)Post-registration
Figure A2.
Illustration of co-registration using a tree trunk from the removal plot. Black: Pre-cut data
and Red: Post-cut data.
Appendix A.3. Additional Results
Result of the frequency distribution of height of different returns for the bi-temporal data of the
control plot, which shows similar results as in Figure 5b, is shown in Figure A3.
(a)26 Feb. 2016
(b)4 Apr. 2016
Figure A3.
Comparison of the frequency distribution with Gaussian kernel density estimation of
different vertical returns with height between 26 February 2016 (
a
) and 4 April 2016 (
b
) data of the
control plot.
Remote Sens. 2018,10, 810 17 of 19
References
1. Crowley, T.J. Causes of climate change over the past 1000 years. Science 2000,289, 270–277.
2.
Woods, P. Effects of logging, drought, and fire on structure and composition of tropical forests in Sabah,
Malaysia. Biotropica 1989,21, 290–298.
3. Wright, S.J. Tropical forests in a changing environment. Trends Ecol. Evol. 2005,20, 553–560.
4.
Schnitzer, S.A.; Bongers, F. Increasing liana abundance and biomass in tropical forests: Emerging patterns
and putative mechanisms. Ecol. Lett. 2011,14, 397–406.
5. Schnitzer, S.A.; Bongers, F. The ecology of lianas and their role in forests. Trends Ecol. Evol. 2002,17, 223–230.
6.
Rodríguez-Ronderos, M.E.; Bohrer, G.; Sanchez-Azofeifa, A.; Powers, J.S.; Schnitzer, S.A. Contribution of
lianas to plant area index and canopy structure in a Panamanian forest. Ecology 2016,97, 3271–3277.
7.
Muthumperumal, C.; Parthasarathy, N. Diversity, distribution and resource values of woody climbers in
tropical forests of southern Eastern Ghats, India. J. For. Res. 2013,24, 365–374.
8.
Bongers, F.; Schnitzer, S.; Traore, D. The importance of lianas and consequences for forest management
in West Africa. BioTerre. 2002. Available online: https://www.researchgate.net/profile/Frans_Bongers/
publication/40797530_The_importance_of_lianas_and_consequences_for_forest_management_in_West_
Africa/links/0c960529614f4e6557000000.pdf (accessed on 22 March 2018).
9. Hegarty, E.E. Distribution and abundance of vines in forest communities. Biol. Vines 1991, 313–334.
10.
Schnitzer, S.A.; Carson, W.P. Treefall gaps and the maintenance of species diversity in a tropical forest.
Ecology 2001,82, 913–919.
11.
Vidal, E.; Johns, J.; Gerwing, J.J.; Barreto, P.; Uhl, C. Vine management for reduced-impact logging in eastern
Amazonia. For. Ecol. Manag. 1997,98, 105–114.
12. Putz, F.E. The natural history of lianas on Barro Colorado Island, Panama. Ecology 1984,65, 1713–1724.
13.
Appanah, S.; Putz, F. Climber abundance in virgin dipterocarp forest and the effect of pre-felling climber
cutting on logging damage [Peninsular Malaysia]. Malays. For. 1984,47, 335–342.
14.
Schnitzer, S.A.; Parren, M.P.; Bongers, F. Recruitment of lianas into logging gaps and the effects of pre-harvest
climber cutting in a lowland forest in Cameroon. For. Ecol. Manag. 2004,190, 87–98.
15.
Pérez-Salicrup, D.R. Effect of liana cutting on tree regeneration in a liana forest in Amazonian Bolivia.
Ecology 2001,82, 389–396.
16.
Gerwing, J.J. Testing liana cutting and controlled burning as silvicultural treatments for a logged forest in
the eastern Amazon. J. Appl. Ecol. 2001,38, 1264–1276.
17.
Marshall, A.R.; Coates, M.A.; Archer, J.; Kivambe, E.; Mnendendo, H.; Mtoka, S.; Mwakisoma, R.; Figueiredo,
R.J.; Njilima, F.M. Liana cutting for restoring tropical forests: A rare palaeotropical trial. Afr. J. Ecol.
2017
,
55, 282–297.
18.
Reid, J.P.; Schnitzer, S.A.; Powers, J.S. Short and long-term soil moisture effects of liana removal in a seasonally
moist tropical forest. PLoS ONE 2015,10, e0141891.
19.
Van Der Heijden, G.M.; Powers, J.S.; Schnitzer, S.A. Lianas reduce carbon accumulation and storage in
tropical forests. Proc. Natl. Acad. Sci. USA 2015,112, 13267–13271.
20.
Hancock, S.; Essery, R.; Reid, T.; Carle, J.; Baxter, R.; Rutter, N.; Huntley, B. Characterising forest gap fraction
with terrestrial lidar and photography: An examination of relative limitations. Agric. For. Meteorol.
2014
,
189, 105–114.
21.
Calders, K.; Origo, N.; Disney, M.; Nightingale, J.; Woodgate, W.; Armston, J.; Lewis, P. Variability
and bias in active and passive ground-based measurements of effective plant, wood and leaf area index.
Agric. For. Meteorol. 2018,252, 231–240.
22.
Danson, F.M.; Hetherington, D.; Morsdorf, F.; Koetz, B.; Allgower, B. Forest canopy gap fraction from
terrestrial laser scanning. IEEE Geosci. Remote Sens. Lett. 2007,4, 157–160.
23.
Jupp, D.L.; Culvenor, D.; Lovell, J.; Newnham, G.; Strahler, A.; Woodcock, C. Estimating forest LAI profiles
and structural parameters using a ground-based laser called ‘Echidna R
.Tree Physiol. 2009,29, 171–181.
24.
Calders, K.; Armston, J.; Newnham, G.; Herold, M.; Goodwin, N. Implications of sensor configuration and
topography on vertical plant profiles derived from terrestrial LiDAR. Agric. For. Meteorol.
2014
,194, 104–117.
25.
Yang, X.; Strahler, A.H.; Schaaf, C.B.; Jupp, D.L.; Yao, T.; Zhao, F.; Wang, Z.; Culvenor, D.S.; Newnham, G.J.;
Lovell, J.L.; et al. Three-dimensional forest reconstruction and structural parameter retrievals using
a terrestrial full-waveform lidar instrument (Echidna R
). Remote Sens. Environ. 2013,135, 36–51.
Remote Sens. 2018,10, 810 18 of 19
26.
Liang, X.; Litkey, P.; Hyyppa, J.; Kaartinen, H.; Vastaranta, M.; Holopainen, M. Automatic stem mapping
using single-scan terrestrial laser scanning. IEEE Trans. Geosci. Remote Sens. 2012,50, 661–670.
27.
Calders, K.; Schenkels, T.; Bartholomeus, H.; Armston, J.; Verbesselt, J.; Herold, M. Monitoring spring
phenology with high temporal resolution terrestrial LiDAR measurements. Agric. For. Meteorol.
2015
,
203, 158–168.
28.
Srinivasan, S.; Popescu, S.C.; Eriksson, M.; Sheridan, R.D.; Ku, N.W. Multi-temporal terrestrial laser scanning
for modeling tree biomass change. For. Ecol. Manag. 2014,318, 304–317.
29.
Liang, X.; Hyyppä, J.; Kaartinen, H.; Holopainen, M.; Melkas, T. Detecting changes in forest structure over
time with bi-temporal terrestrial laser scanning data. ISPRS Int. J. Geo-Inf. 2012,1, 242–255.
30.
Olivier, M.D.; Robert, S.; Fournier, R.A. A method to quantify canopy changes using multi-temporal
terrestrial lidar data: Tree response to surrounding gaps. Agric. For. Meteorol. 2017,237, 184–195.
31.
Parker, G.G.; Harding, D.J.; Berger, M.L. A portable LIDAR system for rapid determination of forest canopy
structure. J. Appl. Ecol. 2004,41, 755–767.
32.
Sánchez-Azofeifa, A.; Portillo-Quintero, C.; Durán, S. Structural effects of liana presence in secondary
tropical dry forests using ground LiDAR. Biogeosci. Discuss. 2015,12, 17153–17175.
33.
Sánchez-Azofeifa, G.A.; Guzmán-Quesada, J.A.; Vega-Araya, M.; Campos-Vargas, C.; Durán, S.M.; D’Souza, N.;
Gianoli, T.; Portillo-Quintero, C.; Sharp, I. Can terrestrial laser scanners (TLSs) and hemispherical photographs
predict tropical dry forest succession with liana abundance? Biogeosciences 2017,14, 977.
34.
Schnitzer, S.A.; Kuzee, M.E.; Bongers, F. Disentangling above-and below-ground competition between lianas
and trees in a tropical forest. J. Ecol. 2005,93, 1115–1125.
35.
Schnitzer, S.A.; Rutishauser, S.; Aguilar, S. Supplemental protocol for liana censuses. For. Ecol. Manag.
2008
,
255, 1044–1049.
36.
Bunting, P.; Armston, J.; Lucas, R.M.; Clewley, D. Sorted pulse data (SPD) library. Part I: A generic file format
for LiDAR data from pulsed laser systems in terrestrial environments. Comput. Geosci. 2013,56, 197–206.
37.
Bunting, P.; Armston, J.; Clewley, D.; Lucas, R.M. Sorted pulse data (SPD) library—Part II: A processing
framework for LiDAR data from pulsed laser systems in terrestrial environments. Comput. Geosci.
2013
,
56, 207–215.
38.
Girardeau-Montaut, D. Cloudcompare-Open Source Project. OpenSource Project. 2011. Available online:
http://www.danielgm.net/cc/ (accessed on 14 November 2016).
39. Massey,F.J., Jr. The Kolmogorov-Smirnov test for goodness of fit. J. Am. Stat. Assoc. 1951,46, 68–78.
40. Burt, A.; Disney, M.; Calders, K. Extracting individual trees from lidar point clouds using treeseg. Methods
Ecol. Evol. 2018, in review.
41.
Popescu, S.C.; Wynne, R.H.; Nelson, R.F. Measuring individual tree crown diameter with lidar and assessing
its influence on estimating forest volume and biomass. Can. J. Remote Sens. 2003,29, 564–577.
42.
Krooks, A.; Kaasalainen, S.; Kankare, V.; Joensuu, M.; Raumonen, P.; Kaasalainen, M. Predicting tree
structure from tree height using terrestrial laser scanning and quantitative structure models. Silva Fenn
2014
,
48, 1125.
43.
Larjavaara, M.; Muller-Landau, H.C. Measuring tree height: A quantitative comparison of two common
field methods in a moist tropical forest. Methods Ecol. Evol. 2013,4, 793–801.
44.
Maas, H.G.; Bienert, A.; Scheller, S.; Keane, E. Automatic forest inventory parameter determination from
terrestrial laser scanner data. Int. J. Remote Sens. 2008,29, 1579–1593.
45.
Calders, K.; Newnham, G.; Burt, A.; Murphy, S.; Raumonen, P.; Herold, M.; Culvenor, D.; Avitabile, V.;
Disney, M.; Armston, J.; et al. Nondestructive estimates of above-ground biomass using terrestrial laser
scanning. Methods Ecol. Evol. 2015,6, 198–208.
46.
Fournier, R.A.; Côté, J.F.; Bourge, F.; Durrieu, S.; Piboule, A.; Béland, M. A method addressing signal
occlusion by scene objects to quantify the 3D distribution of forest components from terrestrial lidar.
In Proceedings of the SilviLaser 2015, La Grande Motte, France, 28–30 September 2015; pp. 29–31.
47.
Hancock, S.; Anderson, K.; Disney, M.; Gaston, K.J. Measurement of fine-spatial-resolution 3D vegetation
structure with airborne waveform lidar: Calibration and validation with voxelised terrestrial lidar.
Remote Sens. Environ. 2017,188, 37–50.
48.
Clark, D.B.; Clark, D.A. Distribution and effects on tree growth of lianas and woody hemiepiphytes in
a Costa Rican tropical wet forest. J. Trop. Ecol. 1990,6, 321–331.
Remote Sens. 2018,10, 810 19 of 19
49.
Avalos, G.; Mulkey, S.S.; Kitajima, K.; Wright, S.J. Colonization strategies of two liana species in a tropical
dry forest canopy. Biotropica 2007,39, 393–399.
50.
Tobin, M.F.; Wright, A.J.; Mangan, S.A.; Schnitzer, S.A. Lianas have a greater competitive effect than trees of
similar biomass on tropical canopy trees. Ecosphere 2012,3, 1–11.
51.
Wilkes, P.; Lau, A.; Disney, M.; Calders, K.; Burt, A.; de Tanago, J.G.; Bartholomeus, H.; Brede, B.; Herold, M.
Data acquisition considerations for Terrestrial Laser Scanning of forest plots. Remote Sens. Environ.
2017
,
196, 140–153.
52.
Zhang, W.; Qi, J.; Wan, P.; Wang, H.; Xie, D.; Wang, X.; Yan, G. An easy-to-use airborne LiDAR data filtering
method based on cloth simulation. Remote Sens. 2016,8, 501.
53.
Besl, P.J.; McKay, N.D. Method for registration of 3-D shapes. In Sensor Fusion IV: Control Paradigms and Data
Structures; International Society for Optics and Photonics: Bellingham, WA, USA, 1992; Volume 1611, pp.
586–607.
c
2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
... Many studies report the prevalence of lianas and their importance in tropical forests. However, few of them utilize quantitative methods to study their structure and their respective impact on an individual tree [17,18]. Recent advances in remote sensing techniques, in particular, Terrestrial Laser Scanning (TLS), provide a great opportunity to study the relationships between tree and liana structures in an unprecedented manner [9,19,20]. ...
... Recent advances in remote sensing techniques, in particular, Terrestrial Laser Scanning (TLS), provide a great opportunity to study the relationships between tree and liana structures in an unprecedented manner [9,19,20]. For instance, Moorthy et al. [17] investigated the potential utilization of TLS in monitoring changes in forest structure after liana removal, demonstrating that TLS could detect local structural changes after liana removal. Furthermore, Bao et al. [21] used TLS data and the Random Forest (RF) to classify lianas and trees with an overall accuracy of 94%. ...
Article
Full-text available
Lianas are self-supporting systems that are increasing their dominance in tropical forests due to climate change. As lianas increase tree mortality and reduce tree growth, one key challenge in ecological remote sensing is the separation of a liana and its host tree using remote sensing techniques. This separation can provide essential insights into how tropical forests respond, from the point of view of ecosystem structure to climate and environmental change. Here, we propose a new machine learning method, derived from Random Forest (RF) and eXtreme Gradient Boosting (XGBoosting) algorithms, to separate lianas and trees using Terrestrial Laser Scanning (TLS) point clouds. We test our method on five tropical dry forest trees with different levels of liana infestation. First, we use a multiple radius search method to define the optimal radius of six geometric features. Second, we compare the performance of RF and XGBoosting algorithms on the classification of lianas and trees. Finally, we evaluate our model against independent data collected by other projects. Our results show that the XGBoosting algorithm achieves an overall accuracy of 0.88 (recall of 0.66), and the RF algorithm has an accuracy of 0.85 (recall of 0.56). Our results also show that the optimal radius method is as accurate as the multiple radius method, with F1 scores of 0.49 and 0.48, respectively. The RF algorithm shows the highest recall of 0.88 on the independent data. Our method provides a new flexible approach to extracting lianas from 3D point clouds, facilitating TLS to support new studies aimed to evaluate the impact of lianas on tree and forest structures using point clouds.
... pnas.org long-term, geographically widespread surveys of lianas and trees to facilitate dynamic hypothesis testing. Also critical are experimental studies that simulate environmental changes, like drought, and terrestrial 3D laser scanners (i.e., LIDAR; 'Laser Imaging, Detection, And Ranging') can improve quantification of liana canopy occupancy and provide detailed forest structure and liana impact assessments (60,61). Contemporarily, the characterization of liana rooting and water uptake strategies, which are currently little explored, presents another promising research direction for model development and understanding tree-liana interactions (11,31,(62)(63)(64). ...
Article
Extending and safeguarding tropical forest ecosystems is critical for combating climate change and biodiversity loss. One of its constituents, lianas, is spreading and increasing in abundance on a global scale. This is particularly concerning as lianas negatively impact forests’ carbon fluxes, dynamics, and overall resilience, potentially exacerbating both crises. While possibly linked to climate-change-induced atmospheric CO 2 elevation and drought intensification, the reasons behind their increasing abundance remain elusive. Prior research shows distinct physiological differences between lianas and trees, but it is unclear whether these differences confer a demographic advantage to lianas with climate change. Guided by extensive datasets collected in Panamanian tropical forests, we developed a tractable model integrating physiology, demography, and epidemiology. Our findings suggest that CO 2 fertilization, a climate change factor promoting forest productivity, gives lianas a demographic advantage. Conversely, factors such as extreme drought generally cause a decrease in liana prevalence. Such a decline in liana prevalence is expected from a physiological point of view because lianas have drought-sensitive traits. However, our analysis underscores the importance of not exclusively relying on physiological processes, as interactions with demographic mechanisms (i.e., the forest structure) can contrast these expectations, causing an increase in lianas with drought. Similarly, our results emphasize that identical physiological responses between lianas and trees still lead to liana increase. Even if lianas exhibit collinear but weaker responses in their performance compared to trees, a temporary liana prevalence increase might manifest driven by the faster response time of lianas imposed by their distinct life-history strategies than trees.
... However, for such understanding to allow for simulating future scenarios would require to be able to simulate how PAVD profile will be impacted by climate change and land use change. The first point is still far from being resolved, while the second point is investigated by an increasing number of studies, particularly regarding logging, forest management, bioregions or biotic interactions (Atikah et al., 2021;Decuyper et al., 2018;Doughty et al., 2023;Felton et al., 2006;Keany et al., 2023;Moorthy et al., 2018). ...
Preprint
Background Vegetation structure is increasingly recognized as a key variable to explain ecosystems states and dynamics. New Remote Sensing tools are available to complement labor intensive field investigations and consider the global biogeography of this parameter. Objectives We propose to model the processes explaining the interaction between vegetation structure and animal community assembly globally, while requiring minimal computing power, based on the most fundamentals assumptions. Methods We integrate spaceborne (GEDI: Global Ecosystem Dynamics Investigation) and ground based (TLS: Terrestrial Laser Scanning) Lidar data in the Madingley general ecosystem model. We compare the outcome of this integration to previous version and to the TetraDensity estimate of animal biomass and Elton traits database for arboreality. Results Animal biomass density simulated by Madingley is closer to global estimates when integrating vegetation structure. The strength of this effect increases with higher cohort body mass and varies with local environmental conditions and stochastic processes. Simulated proportion of arboreality across cohorts is consistently higher than observations. This is consistent with the divergence of biases between model and database. Conclusions Our results concur with our hypotheses about the role of vegetation structure on animal community assembly, as it reduces total animal biomass abundance. However, assessing the accuracy of its relative weight is challenging. While we have global products about arboreality and animal biomass density, they represent modern day ecosystem state, including anthropogenic activity, while Madingley simulates potential ecosystem optimum. Therefore, we call for further research in this field and for challenging modelling attempts to compare with.
... Several studies have shown that TLS is especially well suited for the characterisation of irregularly shaped plants; e.g. lianas (Krishna Moorthy et al., 2018, city trees Wilkes et al., 2018;Kükenbrink et al., 2021), exceptionally tall trees (Disney, 2019;Disney et al., 2020;Demol et al., 2020) Forest aboveground carbon stocks can either be measured destructively, or estimated with ASMs or TLS. Carbon fluxes in and out of forest ecosystems can be measured using the eddy-covariance (EC) technique (Baldocchi, 2003(Baldocchi, , 2014. ...
Thesis
Improving the global monitoring of aboveground biomass (AGB) is crucial for forest management to be effective in global change mitigation and for advancing our understanding of the carbon cycle. Terrestrial laser scanning (TLS) is used to collect extremely precise and detailed 3D point clouds in forest environments. In the last decade, a range of methods have been developed to estimate AGB from TLS data. As such, TLS can address several drawbacks and uncertainties associated with conventional allometric and Earth observation methods that quantify AGB. The objective of this thesis was to evaluate and improve estimates of AGB using TLS. For this, a large suite of TLS and empirical measurements of tree size and mass were collected. Various point cloud processing and tree reconstruction modelling approaches were tested. Wind effects, coregistration inaccuracies and reflectance scattering were major drivers for noise and point cloud inaccuracies, resulting in smaller branches (< 7 cm) to be overestimated in Quantitative Structure Models (QSMs). Stem volumes of broadleaved and coniferous temperate trees were accurately modelled using QSM, but crown volumes were usually overestimated. Several mitigation strategies were developed to correct this overestimation. Inter-tree and stump-to-tip variations in wood basic density (ρ) caused a consistent bias in TLS-derived AGB when measuring ρ at breast height or sourcing it from databases. QSM models were a helpful tool to calculate volume-weighted average ρ and as such constrain inaccuracies. Increment coring and improved knowledge of species-specific stump-to-tip ρ trends are required to obtain accurate AGB from TLS volume. A global synthesis from 10 TLS AGB validation studies featuring 393 trees ranging 13 – 43,000 kg AGB elucidated that TLS-derived AGB of smaller trees (< 1000 kg) was usually overestimated due to scattering and misalignment errors. TLS-derived AGB of big trees (> 1000 kg) on the other hand was nearly unbiased. Conversely, conventional allometric estimates of AGB were more biased than TLS. This thesis contributed to our understanding of the capabilities and current limitations of TLS for acquiring accurate forest AGB estimates. These limitations are mainly technical and computational in nature. Therefore, larger-scale collection of ground-truth data for algorithm benchmarking is needed. TLS will be essential for calibration and validation of several forest biomass Earth observation missions and allometric scaling models.
... Yuan et al. (2019) specifically, based part of their research on some of the main findings of Chapter 2 of this dissertation. On the other hand, using active sensors such as Terrestrial Laser Scanning (TLS), studies have addressed the potential impact of lianas on forest structure (Moorthy et al., 2018;Rodríguez-Ronderos et al., 2016; or attempted to quantify the biomass of lianas and trees (Moorthy et al., 2020. ...
Thesis
Full-text available
Lianas are woody thick-stemmed climbers that use host trees to reach the forest canopy. Studies have shown a remarkable increase in liana abundance in the last two decades, while others have shown that liana abundance is associated with detrimental effects on forest dynamics. Liana abundance presents peaks in highly seasonal forests such as the Tropical Dry Forest (TDF); regions that are under threat for frequent droughts, fires, and anthropogenic pressures. Despite their abundance and relevance in these fragile ecosystems, there are no clear research priorities that help to conduct an efficient detection and monitoring of lianas. This dissertation aims to integrate new remote sensing perspectives to detect and quantify lianas and trees at the TDF. This was addressed using passive (Chapters 2 ‒ 4) and active remote sensing (Chapter 5). Using thermography, Chapters 2 explored the temporal variability of leaf temperature of lianas and trees at the canopy. Temperature observations were conducted in different seasons and ENSO years on lianas and trees infested and non-infested by lianas. The findings revealed that the presence of lianas on trees does not affect the temperature of exposed tree leaves; however, liana leaves tended to be warmer than tree leaves at noon. The results emphasize that lianas are an important biotic factor that can influence canopy temperature, and perhaps, its productivity. Chapter 3 assessed the discrimination of liana and tree leaves using visible-near infrared (VIS-NIR) and longwave infrared (LWIR) spectra. This chapter compared the former contrasting spectral regions, four representations of leaf spectra, twenty-one algorithms of classification, and two contrasting life forms in the context of machine learning to explore the question of whether it is possible to discriminate between liana and tree leaves. The results revealed that both life forms are more accurately discriminated using LWIR spectra (accuracy between 66% and 96%) compared with VIS-NIR spectra (accuracy between 50% and 69%). However, such accuracies of discrimination were achieved depending on the kind of spectral pre-processing and machine learning algorithm. The chapter’s outcomes suggest the possibility to extend the discrimination between lianas and trees to airborne or satellite LWIR observations. The prediction of leaf traits of lianas and trees using Partial Least-Square Regression (PLSR) models based on leaf reflectance or wavelet spectra is addressed in Chapter 4. This chapter revealed that the model performance differs between life forms or between reflectance/wavelet spectra models. Differences in model performance between life forms seemed to be the product of the intraspecific variability of leaf traits within these life forms. Likewise, it was shown that PLSR models based on wavelet spectra help to overcome current limitations of PLSR models based on reflectance spectra. The results showed that the variability of leaf traits between life forms influences predictive models. Thus, the variability of traits between plant groups may have an essential role in estimated errors associated with the mapping of leaf traits. Using Terrestrial Laser Scanning, Chapter 5 evaluated the relationship between fractal geometry and tree-stands metrics on point clouds of trees. The chapter’s results suggested that the intercept extracted from fractal geometry is an accurate and fast parameter that helps predict plant volume, crown coverage, or plant basal area at the tree or stand level. The fractal geometry also revealed that the fractal dimension is strongly associated with the presence/absence of leaves in the point cloud or the number of trees in the stands. Since this method is not susceptible to irregular structures, this method may potentially contribute to quantifying the volume of lianas or buttress roots of trees. Chapter 6 provides future research directions that may help explain the drivers that lead the observed findings or the potential applicability of the results. Overall, this thesis highlighted the need for new efficient and fast approaches that help assess the role and extent of lianas in the tropics. In the absence of a solid understanding of the presence and the effect of lianas in forest dynamics, future predictions of tropical forest productivity will remain speculative.
... Lianas had been detected by an increase in the normalized difference vegetation index (NDVI) and in the photosynthetic vegetation index (PV) in areas where timber harvesting occurs as they are marked by a great number of gaps (Broadbent et al. 2006). New remote sensing technologies such as leaf spectral reflectance and LIDAR scanners has been applied to detect structural vegetation changes related to lianas dominance (Castro-Esau et al. 2004;Sánchez-Azofeifa and Castro-Esau 2006;Sánchez-Azofeifa et al. 2009a, b;Moorthy et al. 2018). Through these leaf distinctions, remote sensing studies start to map and monitor forest areas dominated by lianas, producing large-scale biomass estimates and revealing changes on forest structure across landscapes and biomes (Kalácska et al. 2007;Foster et al. 2008;Ledo et al. 2016;Marvin et al. 2016;Tymen et al. 2016). ...
Article
Full-text available
Key message A systematic review (1950–2018) summarizes the research on woody lianas and their interaction with trees in the Neotropics. We identify knowledge gaps, propose new directions for future studies and discuss the control, management, and conservation of lianas. Abstract Lianas are key components of species composition, structure and dynamics of tropical forests. Current global warming scenario, however, are favoring increases in the abundance and density of lianas in tropical forests, affecting tree growth, fertility, and the number of tree injuries, therefore, increasing tree mortality over time. Here, we present a systematic review of studies on Neotropical lianas and its relation with trees, aiming to (1) establish the current state of ecological research, identifying knowledge gaps and propose new directions and perspectives for future studies; (2) offer baseline knowledge to support the control, management and conservation of lianas. We surveyed the literature on lianas (woody climbers) since 1900 to 2018 retaining 427 papers. We organized the literature by country, vegetation type, topic addressed and whether the study focused exclusively on lianas or lianas and trees. Our review demonstrated the importance of lianas in tropical forests, and the scarcity of studies on woody savannas and especially extremely dry vegetations as the Caatinga seasonally dry forests and xeric shrublands. Regardless of their remarkable importance and their contribution for diversity, biomass and carbon flux, lianas are rarely included in global vegetation models and have been overlooked in restoration, control, and management programs. We must consider the relevance of lianas in maintaining diversity and microclimate, and as resources for native animals, such as pollinators, herbivores, and seed dispersers, as well as for traditional human communities. Research on ecophysiology and functional spectral traits, and management of lianas are among the key areas in the Anthropocene.
... Lianas increase habitat heterogeneity throughout the vertical profile of tropical forests by increasing complexity and structure. For example, measurements using terrestrial and airborne LiDAR, plant area index, and manual leaf harvesting in intact forest and following liana removal experiments have shown that liana stems and leaves add considerable structure and complexity to the understory, mid canopy, and upper canopy of tropical forests (Clark et al. 2008, S anchez-Azofeifa et al. 2009, Marvin et al. 2016, Rodriguez-Ronderos et al. 2016, Moorthy et al. 2018. Liana fruits, flowers, leaves, and stems are also valuable food resources to the animal community. ...
Article
Full-text available
The spatial habitat heterogeneity hypothesis posits that habitat complexity increases the abundance and diversity of species. In tropical forests, lianas add substantial habitat heterogeneity and complexity throughout the vertical forest profile, which may maintain animal abundance and diversity. The effects of lianas on tropical animal communities, however, remain poorly understood. We propose that lianas have a positive effect on animals by enhancing habitat complexity. Lianas may have a particularly strong influence on the forest bird community, providing nesting substrate, protection from predators, and nutrition (food). Understory insectivorous birds, which forage for insects that specialize on lianas, may particularly benefit. Alternatively, it is possible that lianas have a negative effect on forest birds by increasing predator abundances and providing arboreal predators with travel routes with easy access to bird nests. We tested the spatial habitat heterogeneity hypothesis on bird abundance and diversity by removing lianas, thus reducing forest complexity, using a large‐scale experimental approach in a lowland tropical forest in the Republic of Panama. We found that removing lianas decreased total bird abundance by 78.4% and diversity by 77.4% after 8 months, and by 40.0% and 51.7%, respectively, after 20 months. Insectivorous bird abundance and diversity 8 months after liana removal were 91.8% and 89.5% lower, respectively, indicating that lianas positively influence insectivorous birds. The effects of liana removal persisted longer for insectivorous birds than other birds, with 77.3% lower abundance and 76.2% lower diversity after 20 months. Liana removal also altered bird community composition, creating two distinct communities in the control and removal plots, with disproportionate effects on insectivores. Our findings demonstrate that lianas have a strong positive influence on the bird community, particularly for insectivorous birds in the forest understory. Lianas may maintain bird abundance and diversity by increasing habitat complexity, habitat heterogeneity, and resource availability.
... Vertical leaf profiles are an important metric to understand how light is extinguished while reaching the understory. Recent technological developments, such as laser scanning (Krishna Moorthy et al., 2018), have allowed to measure these quantities with greater precision, however, these instruments cannot parse the different components of the leaf profile (i.e. species, growth form). ...
Article
Full-text available
Tropical forests are a critical component of the Earth system, storing half of the global forest carbon stocks and accounting for a third of terrestrial photosynthesis. Lianas are structural parasites that can substantially reduce the carbon sequestration capacity of these forests. Simulations of this peculiar growth form have only recently started and a single vegetation model included lianas so far. In this work we present a new liana implementation within the individual based model Formind. Initial tests indicate high structural realism both horizontal and vertical. In particular, we benchmarked the model against empirical observations of size distribution, mean liana cluster size and vertical leaf distribution for the Paracou site in French Guiana. Our model predicted a reduction of above-ground biomass between 10% for mature stands to 45% for secondary plots upon inclusion of lianas in the simulations. The reduced biomass was the result of a lower productivity due to a combination of lower tree photosynthesis and high liana respiration. We evaluated structural metrics (LAI, basal area, mean tree-height) and carbon fluxes (GPP, respiration) by comparing simulations with and without lianas. At the equilibrium, liana productivity was 1.9t C ha - 1 y - 1 , or 23% of the total GPP and the forest carbon stocks were between 5% and 11% lower in simulations with lianas. We also highlight the main strengths and limitations of this new approach and propose new field measurements to further the understanding of liana ecology in a modelling framework.
... Liana assessment has traditionally required direct measurement by hand, but the immense challenge of conducting fieldwork in liana thickets and the imprecision of allometric equations may be alleviated using terrestrial laser scanning (LiDAR) to estimate 3-D structure beneath the thick mat of vines (Sánchez-Azofeifa et al., 2017;Moorthy et al., 2018). Subsequently, rapidly developing algorithms can facilitate extraction of tree and liana structural attributes from the complex point clouds of LiDAR data . ...
Article
Full-text available
Lianas are woody vines, rooted in the soil, and supported physically by trees. Lianas contribute to forest ecosystem functioning globally, but especially in the tropics and subtropics. However, prolific liana growth following heavy disturbance frequently affects subsequent recovery of forest tree diversity, biomass, structure, and function. Understanding this forest liana dynamic, and its sensitivity to climate and anthropogenic forces, is essential for worldwide forest restoration and climate change mitigation. Here, we synthesise the evidence for both positive and negative effects of lianas on forests and propose a framework that outlines the expected global response of forests to disturbance-induced liana proliferation. Emerging evidence suggests that lianas play a major role in both facilitating and delaying forest recovery following disturbance. At low levels of disturbance and/or where environmental conditions favour tree growth, lianas can facilitate forest recovery by protecting trees from extreme weather, fire, weed invasion and herbivory. However, under conditions where lianas proliferate beyond critical thresholds, positive feedbacks are expected to induce and sustain liana-dominated forest states that can endure for decades or even longer. We conceptualise alternative classes of forest recovery response to disturbance and describe measurement and modelling of liana thresholds.We identify four essential challenges for global change science relating to lianas: (1) incorporation of lianas and sapling stems into forest monitoring and tree stand measurements worldwide; (2) long-term experiments to determine variation in liana-tree competition, and potential drivers across forest successional gradients; (3) identification and prediction of liana thresholds and other alternative forest recovery response classes; and (4) dynamicmechanisticmodelling of forest recovery to determine regional and global variation within and among different recovery response classes, in relation to variation in potential drivers, liana feedbacks and their interactions. Addressing these challenges will determine the importance of lianas in shaping regional and global forest composition, recovery and dynamics.
Article
Full-text available
A preliminary inventory on the above ground biomass and carbon stock of liana associated with landscape-scale fragmentation induced by the Pergau dam along the watershed area of Sungai Long II intake in Jeli, Kelantan was executed. A total of 67 liana individuals (≥1 cm dbh) was enumerated in 0.15 ha of surveyed plot. Lianas with dbh range between 0.06 to 0.95cm were found to be dominant in the inventory site, comprising 41.8% of the proportional abundance followed by 40.3% of lianas with dbh in the range of 1.05 to 1.94cm. The least encountered dbh of lianas were 3.82cm and 4.14cm respectively (1.5%). The total above ground biomass was estimated 37.9 t ha-1 within the inventory site. The contribution of liana biomass to total above ground biomass has been estimated 6.5%, extremely lower than the total above ground biomass contributed by the thirteen host trees (545.3 t ha-1). Carbon sequestration by lianas was estimated 19.8 t C ha-1 , contributing 6.8% in response to level off high CO2 whilst the carbon stock of the host trees was estimated 272.6 t C ha-1. It could be suggested that ecosystems along the Sungai Long II in Pergau, Kelantan have been disturbed and this was based on the high abundance of liana. Although the above ground biomass of lianas was lesser than the host trees, the liana population size should be controlled and in equilibrium within the ecosystem. To ensure the stability of the ecosystem, regular silviculture treatment should be executed in the highly liana infected area.
Article
Full-text available
Recent studies have demonstrated the potential of lidar‐derived methods in plant ecology and forestry. One limitation to these methods is accessing the information content of point clouds, from which tree‐scale metrics can be retrieved. This is currently undertaken through laborious and time‐consuming manual segmentation of tree‐level point clouds from larger‐area point clouds, an effort that is impracticable across thousands of stems. Here, we present treeseg , an open‐source software to automate this task. This method utilises generic point cloud processing techniques including Euclidean clustering, principal component analysis, region‐based segmentation, shape fitting and connectivity testing. This data‐driven approach uses few a priori assumptions of tree architecture, and transferability across lidar instruments is constrained only by data quality requirements. We demonstrate the treeseg algorithm here on data acquired from both a structurally simple open forest and a complex tropical forest. Across these data, we successfully automatically extract 96% and 70% of trees, respectively, with the remainder requiring some straightforward manual segmentation. treeseg allows ready and quick access to tree‐scale information contained in lidar point clouds. treeseg should help contribute to more wide‐scale uptake of lidar‐derived methods to applications ranging from the estimation of carbon stocks through to descriptions of plant form and function.
Article
Full-text available
In situ leaf area index (LAI) measurements are essential to validate widely-used large-area or global LAI products derived, indirectly, from satellite observations. Here, we compare three common and emerging ground-based sensors for rapid LAI characterisation of large areas, namely digital hemispherical photography (DHP), two versions of a widely-used commercial LAI sensor (LiCOR LAI-2000 and 2200), and terrestrial laser scanning (TLS). The comparison is conducted during leaf-on and leaf-off conditions at an unprecedented sample size in a deciduous woodland canopy. The deviation between estimates of these three ground-based instruments yields differences greater than the 5% threshold goal set by the World Meteorological Organization. The variance at sample level is reduced when aggregated to plot scale (1 ha) or site scale (6 ha). TLS shows the lowest relative standard deviation in both leaf-on (11.78%) and leaf-off (13.02%) conditions. Whereas the relative standard deviation of effective plant area index (ePAI) derived from DHP relates closely to TLS in leaf-on conditions, it is as large as 28.14–29.74% for effective wood area index (eWAI) values in leaf-off conditions depending on the thresholding technique that was used. ePAI values of TLS and LAI-2x00 agree best in leaf-on conditions with a concordance correlation coefficient (CCC) of 0.796. In leaf-off conditions, eWAI values derived from DHP with Ridler and Calvard thresholding agrees best with TLS. Sample size analysis using Monte Carlo bootstrapping shows that TLS requires the fewest samples to achieve a precision better than 5% for the mean ± standard deviation. We therefore support earlier studies that suggest that TLS measurements are preferential to measurements from instruments that are dependent on specific illumination conditions. A key issue with validation of indirect estimates of LAI is that the true values are not known. Since we cannot know the true values of LAI, we cannot quantify the accuracy of the measurements. Our radiative transfer simulations show that ePAI estimates are, on average, 27% higher than eLAI estimates. Linear regression indicated a linear relationship between eLAI and ePAI–eWAI (R² = 0.87), with an intercept of 0.552 and suggests that caution is required when using LAI estimates.
Article
Full-text available
The poor constraint of forest Above Ground Biomass (AGB) is responsible, in part, for large uncertainties in modelling future climate scenarios. Terrestrial Laser Scanning (TLS) can be used to derive unbiased and non-destructive estimates of tree structure and volume and can, therefore, be used to address key uncertainties in forest AGB estimates. Here we review our experience of TLS sampling strategies from 27 campaigns conducted over the past 5 years, across tropical and temperate forest plots, where data was captured with a RIEGL VZ-400 laser scanner. The focus is on strategies to derive Geometrical Modelling metrics (e.g. tree volume) over forest plots (≥1 ha) which require the accurate co-registration of 10s to 100s of individual point clouds. We recommend a 10 m × 10 m sampling grid as an approach to produce a point cloud with a uniform point distribution, that can resolve higher order branches (down to a few cm in diameter) towards the top of 30+ m canopies and can be captured in a timely fashion i.e. ∼ 3–6 days per ha. A data acquisition protocol, such as presented here, would facilitate data interoperability and inter-comparison of metrics between instruments/groups, from plot to plot and over time.
Article
Full-text available
Tropical dry forests (TDFs) are ecosystems with long drought periods, a mean temperature of 25 ∘C, a mean annual precipitation that ranges from 900 to 2000 mm, and that possess a high abundance of deciduous species (trees and lianas). What remains of the original extent of TDFs in the Americas remains highly fragmented and at different levels of ecological succession. It is estimated that one of the main fingerprints left by global environmental and climate change in tropical environments is an increase in liana coverage. Lianas are non-structural elements of the forest canopy that eventually kill their host trees. In this paper we evaluate the use of a terrestrial laser scanner (TLS) in combination with hemispherical photographs (HPs) to characterize changes in forest structure as a function of ecological succession and liana abundance. We deployed a TLS and HP system in 28 plots throughout secondary forests of different ages and with different levels of liana abundance. Using a canonical correlation analysis (CCA), we addressed how the VEGNET, a terrestrial laser scanner, and HPs could predict TDF structure. Likewise, using univariate analyses of correlations, we show how the liana abundance could affect the prediction of the forest structure. Our results suggest that TLSs and HPs can predict the differences in the forest structure at different successional stages but that these differences disappear as liana abundance increases. Therefore, in well known ecosystems such as the tropical dry forest of Costa Rica, these biases of prediction could be considered as structural effects of liana presence. This research contributes to the understanding of the potential effects of lianas in secondary dry forests and highlights the role of TLSs combined with HPs in monitoring structural changes in secondary TDFs.
Article
Full-text available
Vegetation structure controls habitat availability, ecosystem services, weather, climate and microclimate, but current landscape scale vegetation maps have lacked details of understorey vegetation and within-canopy structure at resolutions finer than a few tens of metres. In this paper, a novel signal processing method is used to correctly measure 3D voxelised vegetation cover from full-waveform ALS data at 1.5m horizontal and 50cm vertical resolution, including understorey vegetation and within-canopy structure. A new method for calibrating and validating the instrument specific ALS processing using high resolution TLS data is also presented and used to calibrate and validate the ALS derived data products over a wide range of land cover types within a heterogeneous urban area, including woodland, gardens and streets. This showed the method to accurately retrieve voxelised canopy cover maps with less than 0.4% of voxels containing false negatives, 10% of voxels containing false positives and a canopy cover accuracy within voxels of 24%. The method was applied across 100km² and the resulting structure maps were compared to the more widely used discrete return ALS and Gaussian decomposed waveform ALS data products. These products were found to give little information on the within-canopy structure and so are only capable of deriving coarse resolution, plot-scale structure metrics. The detailed 3D canopy maps derived from the new method allow landscape scale ecosystem processes to be examined in more detail than has previously been possible, and the new method reveals details about the canopy understorey, creating opportunities for ecological investigations. The calibration method can be applied to any waveform ALS instrument and processing method. All code used in this paper is freely available online through bitbucket (https://bitbucket.org/StevenHancock/voxel_lidar).
Chapter
The climbing habit in plants has apparently evolved numerous times. Species that climb are well represented in habitats ranging from tropical rain forests through temperate forests to semi-deserts. The Biology of Vines, first published in 1992, is a treatment of what is known about climbing plants, written by a group of experts and covering topics ranging from the biomechanics of twining to silvicultural methods for controlling vine infestations. Also included are detailed accounts of climbing plant evolution, stem anatomy and function, climbing mechanics, carbon and water relations, reproductive ecology, the role of vines in forest communities and their economic importance. The chapters are based on research on herbaceous vines and woody climbers (lianas) in both temperate and tropical zones, deserts and rain-forests and Old and New World areas. Much remains to be learned about the biology of these plants, but this volume provides a substantial foundation upon which further research can be based.