276
classifier to discriminate IAS with relative success, attributed to the differences in
senescence colors between the IAS and the native vegetation. A subsequent study
scaled this approach to satellites, using a combination of Hyperion, Landsat 5, and
aerial photos to define characteristic spectral signatures from 400 to 950 nm for
Chinese tallow (Ramsey III et al. 2005). Pearlstine et al. (2005) also used aerial
photos with larger spatial resolution (37 × 25 m) to map Brazilian pepper tree
(Schinus terebinthifolius) using texture analysis on red, green, and NIR bands to
identify the IAS relatively well.
Multispectral satellite data have been used to map tree IAS with varying levels of
success. Fuller (2005) performed a supervised classification of IKONOS (2 m) and
Landsat ETM+ (30 m) data to detect broad-leafed paperbark (Melaleuca quinquenervia) in Florida; the timing of imagery was chosen to enhance IAS separability.
Cuneo et al. (2009) also used Landsat Enhanced Thematic Mapper (ETM) data to
map African olive (Olea europaea cuspidata) in Australia based on spectral dissimilarity with the native Eucalyptus spp. with an accuracy of 85% and very low confusion between the species. More recently, decadal-scale time series afforded by
sustained land imaging have enabled increased accuracy in cases where phenological cycles can distinguish IAS. Diao and Wang (2016) used a long time series to use
the phenological changes in tamarisk for high-accuracy classification. Hoyos et al.
(2010) mapped glossy privet (Ligustrum lucidum) in Argentina using a time series
of Landsat TM data and machine learning (support vector machines, SVM), achieving classification accuracies of 89%.
Several studies used imaging spectroscopy to map tree IAS (He et al. 2011;
Bradley 2014), e.g., tamarisk (Hamada et al. 2007; Carter et al. 2009), black cherry
(Prunus serotina), black locust (Robinia pseudoacacia) and northern red oak
(Quercus rubra) (Boschetti et al. 2007), Brazilian pepper (Lass and Prather 2004),
and fire tree (Myrica faya) (Asner et al. 2008a, b). The studies determined characteristic IAS spectral profiles (sensu Ramsey III et al. 2005), compared spectral profiles across species using techniques such as SAM (e.g., Lass and Prather 2004), and
correlated them with ground measurements (e.g., Asner et al. 2008a, b).
Lidar in combination with imaging spectroscopy has been found useful for
assessments of tree IAS (Huang and Asner 2009). For example, Asner et al. (2008a,
b) combined imaging spectroscopy and lidar to detect fire tree in Hawaii and measure impacts on forest canopy biochemistry (Fig. 12.3). Hantson et al. (2012)
mapped black cherry and beach rose (Rosa rugosa) in the Netherlands, finding that
the additional height information from lidar improved classification accuracy by
12% over imaging spectroscopy data alone.
Direct detection of IAS on the tree canopy has also been studied. For example,
Cheng et al. (2007) used imaging spectroscopy to detect kudzu (Pueraria montana)
in a pine forest in Western Georgia, United States. They used an MNF transform and
SAM to differentiate the spectral profile of the IAS from the native forest. Wu et al.
(2006) mapped the invasive climbing fern (Lygodium microphyllum) in the Florida
Everglades with a supervised classification of IKONOS imagery to show how it
established in different parts of the forest. Although successful, their results underestimated fern extent in the understory.
E. A. Bolch et al.
classifier to discriminate IAS with relative success, attributed to the differences in
senescence colors between the IAS and the native vegetation. A subsequent study
scaled this approach to satellites, using a combination of Hyperion, Landsat 5, and
aerial photos to define characteristic spectral signatures from 400 to 950 nm for
Chinese tallow (Ramsey III et al. 2005). Pearlstine et al. (2005) also used aerial
photos with larger spatial resolution (37 × 25 m) to map Brazilian pepper tree
(Schinus terebinthifolius) using texture analysis on red, green, and NIR bands to
identify the IAS relatively well.
Multispectral satellite data have been used to map tree IAS with varying levels of
success. Fuller (2005) performed a supervised classification of IKONOS (2 m) and
Landsat ETM+ (30 m) data to detect broad-leafed paperbark (Melaleuca quinquenervia) in Florida; the timing of imagery was chosen to enhance IAS separability.
Cuneo et al. (2009) also used Landsat Enhanced Thematic Mapper (ETM) data to
map African olive (Olea europaea cuspidata) in Australia based on spectral dissimilarity with the native Eucalyptus spp. with an accuracy of 85% and very low confusion between the species. More recently, decadal-scale time series afforded by
sustained land imaging have enabled increased accuracy in cases where phenological cycles can distinguish IAS. Diao and Wang (2016) used a long time series to use
the phenological changes in tamarisk for high-accuracy classification. Hoyos et al.
(2010) mapped glossy privet (Ligustrum lucidum) in Argentina using a time series
of Landsat TM data and machine learning (support vector machines, SVM), achieving classification accuracies of 89%.
Several studies used imaging spectroscopy to map tree IAS (He et al. 2011;
Bradley 2014), e.g., tamarisk (Hamada et al. 2007; Carter et al. 2009), black cherry
(Prunus serotina), black locust (Robinia pseudoacacia) and northern red oak
(Quercus rubra) (Boschetti et al. 2007), Brazilian pepper (Lass and Prather 2004),
and fire tree (Myrica faya) (Asner et al. 2008a, b). The studies determined characteristic IAS spectral profiles (sensu Ramsey III et al. 2005), compared spectral profiles across species using techniques such as SAM (e.g., Lass and Prather 2004), and
correlated them with ground measurements (e.g., Asner et al. 2008a, b).
Lidar in combination with imaging spectroscopy has been found useful for
assessments of tree IAS (Huang and Asner 2009). For example, Asner et al. (2008a,
b) combined imaging spectroscopy and lidar to detect fire tree in Hawaii and measure impacts on forest canopy biochemistry (Fig. 12.3). Hantson et al. (2012)
mapped black cherry and beach rose (Rosa rugosa) in the Netherlands, finding that
the additional height information from lidar improved classification accuracy by
12% over imaging spectroscopy data alone.
Direct detection of IAS on the tree canopy has also been studied. For example,
Cheng et al. (2007) used imaging spectroscopy to detect kudzu (Pueraria montana)
in a pine forest in Western Georgia, United States. They used an MNF transform and
SAM to differentiate the spectral profile of the IAS from the native forest. Wu et al.
(2006) mapped the invasive climbing fern (Lygodium microphyllum) in the Florida
Everglades with a supervised classification of IKONOS imagery to show how it
established in different parts of the forest. Although successful, their results underestimated fern extent in the understory.
E. A. Bolch et al.