271
Three factors make mapping IAS using RS most viable (He et al. 2015). First,
when the IAS is the dominant growth form or has large homogeneous patches, it is
easier to train a classifier to recognize it. For example, water hyacinth is sometimes
the only IAS in lakes, so mapping it is as easy as separating bright green vegetation
from spectrally dark water (Venugopal 2002). This is feasible with simple color IR
aerial photography (Rouse et al. 1975) or multispectral satellite data such as that
from Satellite Pour l’Observation de la Terre (SPOT) or Landsat. Second, when the
target IAS has a unique phenology, it is easier to distinguish from native plants during some parts of the year. For example, Andrew and Ustin (2008) identified perennial pepperweed (Lepidium latifolium) during its flowering period, when it was
spectrally most distinct from the surrounding marsh due to its unique white flowers.
Temporally rich imagery can be used to identify the ideal time period for differentiation along with high spectral resolution to distinguish phenological differences.
Third, the target IAS has a unique chemistry or biophysiology. For example, Khanna
et al. (2011) differentiated water hyacinth from other co-occurring floating aquatic
macrophytes using differences in canopy water content, since water hyacinth is a
succulent with a higher plant-water content than co-occurring species water primrose (Ludwigia peploides) and water pennywort (Hydrocotyle ranunculoides). This
requires a spectrally rich data set that is capable of quantifying canopy biochemistry. These three requirements are well matched with the three domains of RS data:
spatial, temporal, and spectral.
Invasion detection often involves species mapping, which requires much more
data than functional-type or general biodiversity mapping. Often hyperspectral
imagery uses phenology to time the image capture and additional ancillary data
such as altitude are necessary. As mentioned, sensors collect information in three
primary domains: spectral, spatial, and temporal (an additional fourth domain,
radiometric resolution, is critical for aquatic and marine applications – see Sect.
12.2.3 for more details). As a rule of thumb, hyperspectral imagery is rich in data in
the spectral domain, aerial imagery from piloted and unpiloted aircraft in the spatial
domain, and satellite imagery in the time domain. Each of these platforms and sensor types has trade-offs between the three domains and is typically only strong in
one. Selecting the best platform/sensor and fusing the collected imagery with appropriate supplementary data results in the best classification maps. Each species and
habitat presents unique challenges for identifying and mapping IAS using RS,
which we elaborate upon further in the chapter. Regardless of habitat, the general
process of detecting and mapping IAS remains the same and consists of the following steps (see also outlined Fig. 12.2):
1. Identify the target species and/or area. What IAS is affecting biodiversity, ecosystem services, or other economic functions in your area (e.g., transportation)?
What do you know about your target IAS (e.g., spectral characteristics, phenology, ecosystem function, habitat requirements)? Do you know, or can you
hypothesize, the IAS extent and community composition of other species in the
area?
12 Remote Detection of Invasive Alien Species
Three factors make mapping IAS using RS most viable (He et al. 2015). First,
when the IAS is the dominant growth form or has large homogeneous patches, it is
easier to train a classifier to recognize it. For example, water hyacinth is sometimes
the only IAS in lakes, so mapping it is as easy as separating bright green vegetation
from spectrally dark water (Venugopal 2002). This is feasible with simple color IR
aerial photography (Rouse et al. 1975) or multispectral satellite data such as that
from Satellite Pour l’Observation de la Terre (SPOT) or Landsat. Second, when the
target IAS has a unique phenology, it is easier to distinguish from native plants during some parts of the year. For example, Andrew and Ustin (2008) identified perennial pepperweed (Lepidium latifolium) during its flowering period, when it was
spectrally most distinct from the surrounding marsh due to its unique white flowers.
Temporally rich imagery can be used to identify the ideal time period for differentiation along with high spectral resolution to distinguish phenological differences.
Third, the target IAS has a unique chemistry or biophysiology. For example, Khanna
et al. (2011) differentiated water hyacinth from other co-occurring floating aquatic
macrophytes using differences in canopy water content, since water hyacinth is a
succulent with a higher plant-water content than co-occurring species water primrose (Ludwigia peploides) and water pennywort (Hydrocotyle ranunculoides). This
requires a spectrally rich data set that is capable of quantifying canopy biochemistry. These three requirements are well matched with the three domains of RS data:
spatial, temporal, and spectral.
Invasion detection often involves species mapping, which requires much more
data than functional-type or general biodiversity mapping. Often hyperspectral
imagery uses phenology to time the image capture and additional ancillary data
such as altitude are necessary. As mentioned, sensors collect information in three
primary domains: spectral, spatial, and temporal (an additional fourth domain,
radiometric resolution, is critical for aquatic and marine applications – see Sect.
12.2.3 for more details). As a rule of thumb, hyperspectral imagery is rich in data in
the spectral domain, aerial imagery from piloted and unpiloted aircraft in the spatial
domain, and satellite imagery in the time domain. Each of these platforms and sensor types has trade-offs between the three domains and is typically only strong in
one. Selecting the best platform/sensor and fusing the collected imagery with appropriate supplementary data results in the best classification maps. Each species and
habitat presents unique challenges for identifying and mapping IAS using RS,
which we elaborate upon further in the chapter. Regardless of habitat, the general
process of detecting and mapping IAS remains the same and consists of the following steps (see also outlined Fig. 12.2):
1. Identify the target species and/or area. What IAS is affecting biodiversity, ecosystem services, or other economic functions in your area (e.g., transportation)?
What do you know about your target IAS (e.g., spectral characteristics, phenology, ecosystem function, habitat requirements)? Do you know, or can you
hypothesize, the IAS extent and community composition of other species in the
area?
12 Remote Detection of Invasive Alien Species