ICESat/GLAS Canopy Height Sensitivity
Inferred from Airborne Lidar
Craig Mahoney, Chris Hopkinson, Alex Held, Natascha Kljun, and Eva van Gorsel
Abstract
Variations in laser properties and data acquisition times
introduced inconsistencies in Geoscience Laser Altimeter
System (
GLAS
) data. The effect of data inconsistencies, on two
GLAS
height retrieval methods, from three study sites, are in-
vestigated and validated against airborne laser scanning (
ALS
)
percentile heights, from three data sources: all/first return
point clouds, and raster canopy height models.
GLAS
/
ALS
con-
trols were established as a basis against which the influence
of laser number, transmission energy, and seasonality were
assessed through comparison statistics. The favored
GLAS
height method best compared with
ALS
95
th
percentile heights
from an all return point cloud. Optimal
GLAS
data (R
2
= 0.69,
RMSE
= 8.10 m) were noted when
GLAS
acquired data during
summertime from high energy, laser three transmissions. As
GLAS
data can be used in global biomass assessments, there is
a need to understand and quantify the influence of these data
inconsistencies on canopy height estimates.
Introduction
The Ice, Cloud, and Land Elevation satellite (
ICESat
) Geosci-
ence Laser Altimeter System (
GLAS
) offered the first oppor-
tunity to actively monitor land surface parameters describ-
ing terrain and canopy height from an Earth orbit platform
(Rosette
et al.,
2008; Los
et al.,
2012). This system pioneered
waveform Light detection And Ranging (lidar) technology at
regional scales and above, and currently represents the only
near global lidar data set (Fatoyinbo
et al.,
2013; Peterson
et
al.,
2013). These data have been employed across multiple
disciplines, including applications in the cryosphere, atmo-
sphere, oceans, and terrestrial biosphere (Zwally
et al.,
2002;
Spinhirne
et al.,
2005; Dessler
et al.,
2006; Urban
et al.,
2008;
Molijn
et al.,
2011). In the terrestrial biosphere, applications
pertain to achieving sustainable forest management strate-
gies, which requires forest parameters be derived with low
uncertainty (Sun
et al.,
2008; Allouis
et al.,
2013). However,
limitations inherent within
GLAS
data dictate that accurate pa-
rameter retrievals at large scales can be challenging (Nelson,
2010; Rosette
et al.,
2010; Wang
et al.,
2011)
.
In contrast to
GLAS
, airborne laser scanning (
ALS
) data are
captured over smaller areas (10’s to 1000’s km
2
) but with
higher resolution, and typically lower uncertainty. However,
despite differences in scale, data captured from any lidar
system is inherently sensitive to variations in instrument and
data collection configuration, and ground conditions at the
time of acquisition (Hopkinson, 2007; Nelson, 2010). Due to
the relatively large 64 m mean diameter footprint of
GLAS
, it is
generally not possible to extract the location, orientation, and
shape of small objects (e.g., individual tree crowns) that are
readily discernible in
ALS
data. The suitability of
ALS
as a bio-
sphere remote sensing tool has been demonstrated in multiple
projects, aimed at simulating vegetation parameters such as:
canopy height (Dubayah
et al.,
2000; Means
et al.,
2000; Lefsky
et al.,
2002), gap fraction or fractional cover (Lefsky
et al.
, 1999;
Hopkinson
et al.,
2009), stand volumes, biomass, basal area,
leaf area index (
LAI
), and diameter at breast height (
DBH
) (Lefsky
et al.
, 1999; Lefsky
et al
., 1999; Means
et al
., 1999; Naesset,
2002; Clark
et al.
, 2004; Patenaude
et al
., 2004; Ni-Meister
et al
.,
2010; Rosette
et al
., 2011). Many canopy or stand attributes are
derived as a function of height (or its percentile representation),
with high accuracy when compared to or calibrated from field
data (Means
et al.,
1999; Means
et al.,
2000). The continued op-
erational use of
ALS
data as a means to model canopy attributes
justifies its use as “truth” against which
GLAS
canopy height es-
timates can be compared. However, which of the available
GLAS
or
ALS
height estimates are most suited for comparison and
further use in canopy attribute models are not
a priori
known
.
Variations in canopy height associated with
GLAS
’s system
setup originate from discrete, sequential operation periods,
approximately one month long, two to three times per year
(known as laser campaigns). Within each campaign, the spatial
energy distribution of footprints was noted to vary, suggested to
be due to variations in the spatial distribution of light from the
laser diode pump arrays with time (Abshire
et al.
, 2005; Sirota
et al
., 2005). Consequently,
GLAS
footprint dimensions (semi-
major and semi-minor axes) varied with each laser campaign,
and as a result, the consistency in
GLAS
canopy height estimates
between lasers, and possibly laser campaigns, is uncertain
.
Each laser was designed to fire with a mean transmit energy
between 65 mJ and 80 mJ (Abshire
et al.
, 2005). However, un-
predictable energy losses were observed in each laser as a func-
tion of operating time, the resultant of suspected component
failure(s). Each laser lost energy at an accelerated rate to the
point where laser 1 failed after a single campaign. As a result,
operating strategies for lasers 2 and 3 were revised, extending
their operational life until they eventually failed. The effect of
laser energy decay on
GLAS
derived canopy height is unknown.
The temporal separation between each one-month long la-
ser campaign creates difficulties in assessing the consistency
of
GLAS
canopy height estimates between varying phenologi-
cal conditions, where phenological condition refers to the
seasonal growth state of an ecosystem at the time of
GLAS
data
acquisition. In Australia it is atypical for most vegetation to
fully defoliate, as a less extreme phenological cycle operates
Craig Mahoney and Chris Hopkinson are with the Department
of Geography, University of Lethbridge, Lethbridge, AB T1K
3M4
;
.
Alex Held is with CSIRO Land and Water, Christian Road,
Acton ACT 2601, Australia. (
)
Natascha Kljun is with the Department of Geography, Swansea
University, Singleton Park, Swansea, SA2 8PP, UK.
(
)
Eva van Gorsel is with CSIRO Oceans and Atmosphere,
Wilf Crane Crescent, Yarralumla ACT 2600, Australia. (eva.
)
Photogrammetric Engineering & Remote Sensing
Vol. 82, No. 5, May 2016, pp. 351–363.
0099-1112/16/351–363
© 2016 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.82.5.351
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2016
351
