Vertical Structure Complexity Assessment of Tropical Forests from a Portable LiDAR System

1

• Vertical Structure Complexity Assessment of
  Tropical Forests from a Portable LiDAR System
• Amanda Cooper1, John Weishampel1, Jason Drake2,
  David Clark3, Geoffrey Parker4
• Department of Biology, The University of Central
  Florida 2. U.S Forest Service, Tallahassee,
  Florida, 3. La Selva Biological Station, Costa
  Rica, 4. Smithsonian Environmental Research
  Center
• Contact Author acooper_at_pegasus.cc.ucf.edu

Introduction Forests spatial organization is
unique compared to other ecosystems because it
incorporates a significant vertical component.
This vertical component (often denoted as
vertical structure) plays a critical role in
ecosystem functions, controls microclimate
variation, provides habitat for interstitial
species, and maintains terrestrial stocks of
carbon. Tropical moist forests are among the
most biologically diverse ecosystems, many
regions being considered biodiversity hot spots.
Tropical forests also contain the largest stores
of terrestrial carbon of all forest systems on
earth, making decreases above-ground biomass
within tropical forest significant to the global
carbon budget. Localized and broad-scale
changes to tropical forest vertical structure are
becoming a prevalent trend, mainly due to logging
practices (Nepstad 1999) and global climate
change (Clark and Clark, 1994). There is growing
need to identify altered tropical forests
structure across the globe as changes are
expected to have a profound impact on
biodiversity and the global carbon cycle. Field
derived measurements, such as foliage height
diversity introduced by MacArthur and Horn
(1969), is effective in describing forest
structure but difficult to implement on a
regional or continental scale. Interest in
broad-scale assessment of forest structure has
driven the exploration of remote sensing
technology directed at evaluating forest
structure. LiDAR (Light Detection and Ranging)
remote sensing depicts three-dimensional surface
features enabling accurate estimations of ground
elevation, vertical forest structure, and canopy
topography. LiDAR has successfully mapped
above-ground biomass in tropical rain forest
(Drake et al. 2002), demonstrating potential
regional level evaluation of forest structure.
Currently, satellite LiDAR (IceSAT/GLAS) does not
provide a sufficient resolution for forest
evaluation. Airborne LiDAR (LVIS, EAARL) systems
have been most effective in forest applications
but geographic coverage is limited due to the
cost of airborne missions. This research
evaluates a ground based system, SYCLPS
(Structure Yielding Canopy LiDAR Portable
System), that may provide comparable information
to airborne or satellite data from the field.
SYCLPS utilizes a first return, upward facing
LiDAR to provide distributional information of
forest structure. This is a preliminary
assessment SYCLPS ability to differentiate
between primary and secondary tropical forests.
We expect that primary forest will demonstrate a
higher diversity in canopy structure layers over
the secondary forest, demonstrating a means for
distinguishing between the two. Data were
collected at La Selva Biological Station as both
primary and secondary forest patches are present
within the station. Additionally, field level
and LVIS LiDAR data have been collected
throughout La Selva, providing a basis for future
comparisons.
Abstract Vertical structure is an important
physical attribute of a forest, influencing the
microclimate, biogeochemical cycling, and
biodiversity. Tropical forests have a highly
complex structure that is altered by both natural
and anthropogenic disturbances. Such
disturbances could permanently affect the
abundance of biodiversity these forest support.
Current field methods for quantifying vertical
structure include field-based forest survey
methods which utilize indicator values such as
stem density and dbh (diameter at breast height)
and labor intensive optical point quadrate
methods that maybe inconsistently interpreted.
LiDAR (Light Detection And Ranging) remote
sensing provides a method for surveying forest
structure that is repeatable and less exhaustive
for researchers. LiDAR data for forests are
primarily collected via airborne (e.g. LVIS,
EAARL) and occasionally via satellite (e.g. GLAS)
platforms. Satellite-based LiDAR is still
lacking at moderate resolutions and airborne
LiDAR has only been collected in a few
broad-scale studies because of the costs of data
collection. This research focuses on the use of
a portable LiDAR system for tropical forest
survey. Our system, SYCLPS (Structure Yielding
Canopy LiDAR Portable System), utilizes a first
return, upward facing LiDAR (Riegl
LD90-3100VHS-FLP) to provide distributional
information of the canopy components. Surveys at
the La Selva Biological Station, Costa Rica in
July 2005 demonstrate that SYCLPS is a useful
tool for defining canopy vertical structure.
SYCLPS data were able to highlight differences in
canopy organization between primary and secondary
forests at La Selva. Further work will use
SYCLPS to develop pseudo-waveforms to mimic
large-footprint sensors and extend the use of
SYCLPS into forest management applications.
Results Figure 4 shows the average frequency
distribution of raw height return values for the
primary and secondary forest transects. As
expected in tropical forests, the distribution is
heavily skewed towards the lowest canopy
components. The majority of measurements were
found in the lowest 5 meters of the canopy.
Field observations suggest that the primary
forest had a denser under story which might block
the upper levels, making the upper canopy less
apparent in the SYCLPS data compared to secondary
forest. The MacArthur-Horn transform can
correct for this obstruction of the upper canopy
canopy structure using a log correction factor.
A comparison between the raw and MacArthur-Horn
transformed data is provided in Figure 5. The
transformed data reveal a much more diverse
upper-level canopy. The transformed primary
forest appears to be fairly even in distribution
within the upper canopy, while the secondary
forest appears to follow more of a binomial
distribution with a peak in the upper canopy and
in the lower canopy. The 95, 75, 50 and 25
quartiles are shown in Figure 6 and though the
secondary forest responses are located at
slightly taller height intervals than the primary
forest, the distributions are approximately the
same. The primary and secondary forest
distributions were not found statistically
different using the Kolmogorov-Smirnov test.
More promising results are The average canopy
height diversity (Fig. 7) and canopy height
evenness (Fig. 8) indices for primary and
secondary forests. These values were found to be
statistically different using an independent t
test. The primary forest plots were found to be
more diverse and more evenly distributed than the
secondary forest plots. This result was expected
based on field based observations at each forest
plot and what is known of tropical primary and
secondary forests.
Conclusion The results demonstrate the
potential of SYCLPS as a tool for evaluating
forest structure, though more consideration is
needed in the sampling design of forest
transects. In this study, fourteen primary and
five secondary forests were surveyed. Several
other sites were visited but were either too over
grown at the ground level or had slopes too high
to safely and accurately survey with the
front-hanging SYCLPS. In order to survey sites
encompassing the full range of forest gap-phase
and topography, a new frame for SYCLPS has been
designed. The new SYCLPS frame design is now in
a back pack (Figure 9), allowing the researcher
to see their path and penetrate the vegetation
before the laser (Figure 10) Further
investigation is needed to determine the
sufficient sample size required to capture the
full variability of primary and secondary forest
with SYCLPS. Comparisons between SYCLPS and the
Towers vertical transect data and airborne LVIS
data will also be investigated to determine the
utility of SYCLPS for tropical forest management
and the potential as a tool for understanding
large footprint, potentially satellite based
waveform data.
Methods This study integrates field-based and
remotely sensed data collection. SYCLPS (Fig. 1)
consists of a LD90-30VHS-FLP range finder
manufactured by Riegl, a field ruggedized laptop
computer, and the front hanging, aluminum frame
modeled after Parker et al. (2004). SYCLPS
collects discrete return distances of
interception points within the canopy at 1/1200
of a second, providing high spatial resolution
data that relate to the vertical distribution of
canopy components. It has a maximum range of 300
meters high intensity measurement settings. The
beam width of the sensor has been measured to
reach an oblique area of 25.6 cm2 at a height of
50 meters (Parker et al. 2004), which is expected
to be sufficiently small for penetrating gaps in
a tropical canopy. SYCLPS measurements were
obtained at La Selva Biological Station in July
2005 in conjunction with the NSF funded TOWERS
project headed by Dr. Steve Oberbauer and Dr.
David Clark. The TOWERS project goals are to
collect fine resolution vertical measurements of
foliage and biomass and to measure their
biochemical components. The TOWERS projects
tropical rain forest vertical structure transect
sites have been selected over a stratified random
sample of soil type, soil nutrients, slope angle
and land use history (Clark Clark 2000) so that
data can be interpolated to landscape level.
Each vertical tower consisted of cube intervals 2
x 2 x 1.87 meters projecting up to the canopy
from which canopy components (leafs, branches,
lianas) were collected. SYCLPS measurements of
vertical organization were collected along four
10-meter transects radiating in four cardinal
directions from the vertical tower site (Fig. 2).
Fourteen primary and five secondary forest sites
were surveyed in total as topography and under
story growth proved to be to difficult to
traverse at some of the sites (Fig. 3). These
difficulties have lead to refinement of the
SYCLPS design (see Conclusions). Raw data
output from SYCLPS requires several preprocessing
steps to derive the vertical height distribution.
Preprocessing includes, removal of no data
values which is determined to be either lt0.2
meters or sky, and standardization of data
collected in 1 meter horizontal increments along
each ten meter transect. This is done to account
for sampling biases between sections,
facilitating intra- and inter-transect
comparison. Once normalized, the data can be put
into vertical bins to provide frequency
distributions for each transect. The
MacArthur-Horn transform (MacArthur and Horn,
1969) was applied to these vertical distributions
to correct for occlusion of the upper canopy from
the dense understory. From the transformed
profiles, height diversity and evenness indices
can also be calculated to determine the diversity
of the canopy structure. The Shannon-Weiner
Index was calculated for primary and secondary
forest plots to determine if primary forests were
actually more structurally diverse than secondary
forests. (Fukushima, 1998)
References Clark, D.B. and D.A. Clark. 1994.
Climate-induced annual variation in canopy tree
growth in a Costa Rican tropical rain forest.
Journal of Ecology. 82865-872 Drake, J.B., R.O.
Dubayah, D.B. Clark, et al. 2002. Estimation of
tropical forest structural characteristics using
large-footprint LiDAR. Remote Sensing of
Environment. 79305-319. Fukushima, Y., T Hiura,
S. Tanabe. 1998. Accuracy of the Mac-Arthur Horn
method for estimating a foliage profile.
Agriculture and Forest Meteorology. 92203-210.
MacArthur, R.H., and H.S. Horn. 1969. Foliage
Profile by Vertical Measurements. Ecology.
53749-752 Nepstad, D.C., A.Verissimo, A.
Alencar, et al. 1999. Large-scale impoverishment
of Amazonian forests by logging and fire.
Nature. 398505-508. Parker, G.G., D. Harding, M.
Berger. 2004. A Portable LIDAR System for Rapid
Determination of Forest Canopy Structure. Journal
of Applied Ecology. 41755-767.