Filtering Global Land and Surface Altimetry Data
(GLA14) for Elevation Accuracy Determination
Jean-Samuel Proulx-Bourque, Ramata Magagi, and Norman T. O’Neill
Abstract
This paper presents a filtering method for
ICES
at
Global Land
Surface Altimetry data (
GLA14
), which is based on indicators
to detect potentially contaminated
GLA14
elevation points.
Potential contamination sources include attitude miscalcula-
tion, saturated echoes, equipment noise, the atmosphere, and
variable elevation within footprints. For a study site located
in Northern Canada, this multi-indicator filter provided a 19
percent reduction in the root mean square error for elevation,
when compared to Canadian Digital Elevation Data (
CDED
).
This result demonstrates the method’s ability to provide an
improved dataset for vertical accuracy evaluation, with re-
spect to unfiltered
GLA14
data. The improvement was achieved
with a rejection rate of 69 percent. However, due to the high
density of the unfiltered
GLA14
data over the study site, a spa-
tially homogeneous distribution of elevation points was main-
tained, even after filtering. Results also showed the rejection
efficiency of most indicators, as well as their complementarity.
Introduction
Many fields of science today rely on good elevation informa-
tion. A Digital Elevation Model (
DEM
) can provide this infor-
mation but may be subject to uncertainties. Depending on the
application, these uncertainties may greatly affect the ex-
pected results and derived conclusions. To use elevation data
appropriately, its quality, or specifically its accuracy, must be
known. The accuracy of a
DEM
can be calculated using precise
reference points (
NDEP
, 2004). For Northern Canada, the Aerial
Survey Database (ASDB), composed of aerotriangulated pho-
togrammetric reference points, has been traditionally used for
this purpose but is limited in terms of density and availability
(Beaulieu and Clavet, 2009). The Global Land and Surface
Altimetry Data (
GLA14
) has sufficient coverage, density and
accuracy to be used as an alternative (Beaulieu and Clavet,
2009). However, contamination sources can affect the signal
and lead to erroneous elevation values (Brenner
et al
., 2007).
Elevation accuracy evaluation is improved by the removal
of inaccurate points from the reference dataset. Different ap-
proaches have been used for this purpose: removing the outli-
ers (Beaulieu and Clavet, 2009; Carabajal
et al.
, 2009; Toutin
et al.
, 2013); using thresholds on physical parameters of the
transmitted or echoed pulses (Brenner
et al.
, 2007; Zwally
et
al.
, 2008); and using statistics (Huang
et al.
, 2013). While the
aim of the aforementioned studies was to evaluate the eleva-
tion accuracy, this objective was only partially fulfilled since
these authors only dealt with some contamination sources.
The method presented in this paper relied on a combi-
nation of these approaches to remove all elevation points
potentially affected by
GLA14
contamination sources. To
achieve this objective, sources of signal contamination affect-
ing
GLA14
data were identified. Then, for each contamination
source, corresponding indicators were selected to discrimi-
nate between contaminated and non-contaminated elevation
values.
GLA14
data were then filtered and each of the selected
indicators was evaluated based on the improvement of the
elevation accuracy of a Canadian Digital Elevation Data (
CDED
)
dataset. All indicators were then combined, and the perfor-
mance of the filtering method was evaluated in the same way.
Study Site and Data
Study Site
The Area Of Interest (
AOI
) covers a 23,000 km² area located in
the Northwest Territories of Canada (Figure 1). Circa-2000 data
shows that about 70 percent of the land cover consists of bare
land, water, and low-lying vegetation. The remaining 30 percent
is covered with isolated tree patches and shrubs. The relief is
relatively flat with almost all slopes (calculated from the
CDED
)
ranging between 0 and 10 percent grade. Based on historical cli-
mate data, snow cover is generally present from October to April.
Canadian Digital Elevation Data (CDED)
The
CDED
provides complete coverage of Canada’s elevation in
raster format (Geobase, 2014). Elevation values are orthomet-
ric, with respect to the Earth Gravitational Model 1996 (
EGM96
).
Planimetric coordinates are geographic and referenced to the
North American Datum of 1983 (
NAD83
). Where better data
were not available,
CDED
s were generated from the interpola-
tion of contours (Beaulieu and Clavet, 2009). This is the case at
the location of our study site, where the pixel spacing is 20 m.
Global Land Surface Altimetry (GLA14) Data
The Geoscience Laser Altimeter System (
GLAS
) is part of the
Ice, Cloud and land Elevation Satellite (
ICES
at
) mission, which
was operational from 2003 to 2009 (
NSIDC
, 2014a). The altim-
eter’s laser beam, with a wavelength of 1064 nm, produces
footprints of ~70 m diameter on the ground every ~170 m
along the orbit (Brenner
et al.
, 2007). Elevation values are
calculated using the two-way displacement time of the pulse
between the target and the sensor (Schutz, 2002).
GLAS
data
are distributed as different standard products, with different
levels of processing, for different purposes. The
GLA14
(Global
Land Surface Altimetry Data) product provides elevation val-
ues for land surfaces. For this product, the centroid of the last
peak is used to compute the range and elevation (Brenner
et
al.
, 2011).
GLA14
data are referenced to the TOPEX/Poseidon
ellipsoid. Compared to the
WGS84
ellipsoid, elevation values
are 70 cm below and planimetric coordinates are considered
Jean-Samuel Proulx-Bourque is with the Université de
Sherbrooke & Natural Resources Canada (NRCan), 2144-010
King West Street, Sherbrooke (Quebec), Canada J1J 2E8 (Jean-
).
Ramata Magagi and Norman T. O’Neil are with the Université
de Sherbrooke, Département de géomatique appliquée -
Faculté des lettres et sciences humaines, 2500, boulevard de
l’Université, Sherbrooke, (Québec), J1K 2R1.
Photogrammetric Engineering & Remote Sensing
Vol. 81, No. 9, September 2015, pp. 701–707.
0099-1112/15/701–707
© 2015 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.81.9.701
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
September 2015
701
