PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
March 2014
217
Comparison of Simulated HyspIRI with
Two Multispectral Sensors for Invasive
Species Mapping
Aaryn D. Olsson and Jeffrey T. Morisette
Abstract
This paper assesses the potential of a single
H
YSP
IRI
scene to
estimate cover of the non-native invasive buffelgrass (Pennise-
tum ciliare) in a heterogeneous Sonoran Desert scrub ecosys-
tem. We simulated
H
YSP
IRI
(60m) along with two multispectral
sensors, Thematic Mapper (
TM
; 30m) and Advanced Space-
borne Thermal Emission and Reflection Spectrometer (
ASTER
;
15m), from high-resolution Airborne Visible/Infrared Imaging
Spectrometer (
AVIRIS
; 3.2m) imagery in an area infested by
buffelgrass near Tucson, Arizona. We compared classification
accuracies of all simulated sensors at spatial resolutions of
15m, 30m, and 60m to evaluate tradeoffs of spectral and spa-
tial resolution across the sensors. Although spectroscopically
superior to Landsat
TM
and
ASTER
,
ASTER
easily outperformed
H
YSP
IRI
for small infestations (225m
2
) on account of its spatial
resolution. Shortwave-infrared bands near 2.2µm were key
indicators for both
H
YSP
IRI
and
ASTER
, highlighting the benefit
of narrow-wave
SWIR
for mapping invasive species in arid
ecosystems.
Introduction
Invasive species pose a rising global ecosystem challenge due
to impacts to ecosystem structure, function, diversity, nutrient
cycling, and disturbance regimes (Mooney and Hobbs, 2000).
Managers and researchers require readily available data and
tools for mapping and monitoring invasive species, yet the
readily available satellite data are often too coarse spatially,
spectrally, temporally, or a combination thereof to effectively
and consistently map invasive species (Turner
et al.,
2003).
While deca-resolution (Morisette, 2010) broadband multispec-
tral satellite imagery with regular return intervals such as the
Landsat and Satellite Pour l’Observation de la Terre (
SPOT
)
families of sensors have been widely successful at mapping
vegetation communities and land-cover change, mapping
individual plant species has been limited to cases in which
large continuous areas have become invaded (e.g., Peterson,
2005; Bradley and Mustard, 2006) and typically requires dis-
tinct phenological differences between natives and invaders
(Huang and Asner, 2009). Many would argue that invasions
that reach this scale are either already unmanageable, or their
presence is already well known. Regardless, small popula-
tions can play a disproportionate role in the rate of spread and
treatment efficacy of invasions (Moody and Mack, 1988; Frid
Aaryn D. Olsson is with Northern Arizona University, 1298
S. Knoles, Dr., Flagstaff, AZ 86011 (
).
Jeffrey T. Morisette is with the US Geologic Survey, Fort
Collins Science Center, 2150 Centre Ave Bldg C, Fort Collins,
CO 80526.
Photogrammetric Engineering & Remote Sensing
Vol. 80, No. 3, March 2014, pp. 217–227.
0099-1112/14/8003–217/$3.00/0
© 2014 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.80.3.217
and Wilmshurst, 2009). Thus, mapping small populations
remains a critical need for resource managers facing plant
invasions.
A key limitation of multispectral imaging in dryland
ecosystems has been an inability to consistently discriminate
between mineral soil and non-photosynthetic vegetation (
NPV
)
(Olsson
et al
., 2011). Mineral soil and
NPV
have distinct wave-
length features in the
SWIR
, particularly in the 2.0 to 2.4µm
region (Asner and Lobell, 2000; Nagler
et al
., 2003). The
multispectral
SWIR
instrument on
ASTER
promised to provide
this capacity, but unfortunately, it suffered a series of failures
that first limited its utility (Iwasaki and Tonooka, 2005) and
ultimately rendered the
SWIR
instrument completely inoper-
able (ASTER Science Office, 2009).
Invasive species mapping has been more successful when
hyperspectral imagery has been used in classification or target
detection.
AVIRIS
and Hyperion have been used to map inva-
sions (Ustin
et al
., 2001; Underwood
et al
., 2003; Lass
et al
.,
2005; Pengra
et al
., 2007; Asner
et al
., 2008), yet both have
their drawbacks.
AVIRIS
is an airborne sensor with 224 contig-
uous band channels measuring upwelling radiance at 0.01µm
intervals between 0.40 and 2.50µm. As an airborne sensor it
can be flown at varying altitudes, resulting in spatial resolu-
tions between 2 and 20 meters. Commissioning
AVIRIS
requires
a research experiment proposal and is effectively limited to a
small number of projects and locations. Hyperion is an experi-
mental sensor mounted on the
NASA
EO-1 satellite and has
220 bands measuring radiance at 0.01µm intervals between
0.40 and 2.50µm at a 30m spatial resolution. While Hyperion
can visit the same location on a regular basis (every five to ten
days with off-nadir pointing), there is a very limited collec-
tion cycle due to its narrow swath and its limited lifetime. As
such, spatial coverage and temporal resolution from either of
these two sensors are limited. Also, as an experimental instru-
ment, Hyperion has issues with cross-track calibration and
low signal-to-noise ratio (
SNR
), particularly in the shortwave-
infrared (Datt
et al
., 2003; Kruse
et al
., 2003). All of these
issues limit the usefulness of these two national assets to con-
tribute to any operational invasive species detection program.
H
YSP
IRI
(Hyperspectral InfraRed Imager) is one of
NASA
’s
“decadal survey” missions that aims to address these and
other monitoring concerns (NAS, 2007).
H
YSP
IRI
is planned to
have an imaging spectrometer that measures upwelling radi-
ance from the visible to short wave infrared (
VSWIR
: 0.38µm
