386
(Justice and Townshend 1981). In RS studies, ground data still are mainly collected
for accuracy assessments or the validation of map products. Most often, land cover
or vegetation maps are produced through image interpretation (Bartholomé and
Belward 2005; Bicheron et al. 2008). Although the importance of accuracy assessments has been pointed out in the literature (Stehman 2001; Justice and Townshend
1981; Johannsen and Daughtry 2009), validation of map products usually means
using high-resolution RS images to validate coarser resolution maps (Congalton
et al. 2014). Map validations are often based on agreement among random points
(Stehman 2001)—i.e., the extent to which land cover classes that random points fall
into match the investigator’s interpretation of land cover visible in an image. While
this procedure makes sense for global map products distinguishing few vegetation
classes, local map products clearly benefit from the collection of ground data for
validation. However, the importance of ground reference data goes far beyond map
accuracy assessments. In fact, ground reference data are essential for remote sensing of plant biodiversity. Collecting ground reference data during spectral field campaigns provides a great opportunity to bridge the gap between RS science and
ecology, two fields that are uniquely positioned to together develop methods to
assess biodiversity across large spatial scales, continuously and in a detailed way.
These assessments are needed to provide information about the current status of
ecosystems; to predict the distribution of biodiversity, ecosystem function, and ecosystem processes into the future; and to counteract detrimental changes in ecosystems associated with global change. One reason to advocate for field campaigns is
that remotely sensed images, which provide information pixel by pixel, always
obscure part of the information on the ground, with the amount of hidden information depending on pixel size (Atkinson 1999). In order to understand the information provided by remotely sensed images of vegetation, it is critical to study the
spectral characteristics of plants, their links to plant traits, and their influence on
ecosystem properties at the sub-pixel level, because spectral variation is progressively lost when spectra of individual plants and non-vegetation features blend
together at increasing spatial resolutions (Atkinson 1999).
This chapter deals mainly with planning field work and the collection of vegetation spectra with field spectrometers on the ground, which can subsequently be
linked to other ecological data and/or RS data to investigate biological phenomena.
Data collection for airborne spectroscopy is discussed as well, while other RS methods such as unmanned aerial systems (UASs), towers, and trams are covered in
more detail in Gamon et al. (Chap. 16). Focus is also placed on data organization
and management, particularly because these aspects of planning tend to receive less
attention than, e.g., planning of sample collection, yet they are critical to a successful field campaign.
This chapter was written in full awareness that “good practices” are everevolving. The relative importance of, and acquisition methods for, ground data,
including ecological data, depends on the research question, on the project goals, as
well as on study scale, spectroscopic methods and RS data used, budget, time, site
accessibility, and the personnel and their training (Justice and Townshend 1981).
Likewise, spectral processing standards evolve and software goes out of date
A. K. Schweiger
(Justice and Townshend 1981). In RS studies, ground data still are mainly collected
for accuracy assessments or the validation of map products. Most often, land cover
or vegetation maps are produced through image interpretation (Bartholomé and
Belward 2005; Bicheron et al. 2008). Although the importance of accuracy assessments has been pointed out in the literature (Stehman 2001; Justice and Townshend
1981; Johannsen and Daughtry 2009), validation of map products usually means
using high-resolution RS images to validate coarser resolution maps (Congalton
et al. 2014). Map validations are often based on agreement among random points
(Stehman 2001)—i.e., the extent to which land cover classes that random points fall
into match the investigator’s interpretation of land cover visible in an image. While
this procedure makes sense for global map products distinguishing few vegetation
classes, local map products clearly benefit from the collection of ground data for
validation. However, the importance of ground reference data goes far beyond map
accuracy assessments. In fact, ground reference data are essential for remote sensing of plant biodiversity. Collecting ground reference data during spectral field campaigns provides a great opportunity to bridge the gap between RS science and
ecology, two fields that are uniquely positioned to together develop methods to
assess biodiversity across large spatial scales, continuously and in a detailed way.
These assessments are needed to provide information about the current status of
ecosystems; to predict the distribution of biodiversity, ecosystem function, and ecosystem processes into the future; and to counteract detrimental changes in ecosystems associated with global change. One reason to advocate for field campaigns is
that remotely sensed images, which provide information pixel by pixel, always
obscure part of the information on the ground, with the amount of hidden information depending on pixel size (Atkinson 1999). In order to understand the information provided by remotely sensed images of vegetation, it is critical to study the
spectral characteristics of plants, their links to plant traits, and their influence on
ecosystem properties at the sub-pixel level, because spectral variation is progressively lost when spectra of individual plants and non-vegetation features blend
together at increasing spatial resolutions (Atkinson 1999).
This chapter deals mainly with planning field work and the collection of vegetation spectra with field spectrometers on the ground, which can subsequently be
linked to other ecological data and/or RS data to investigate biological phenomena.
Data collection for airborne spectroscopy is discussed as well, while other RS methods such as unmanned aerial systems (UASs), towers, and trams are covered in
more detail in Gamon et al. (Chap. 16). Focus is also placed on data organization
and management, particularly because these aspects of planning tend to receive less
attention than, e.g., planning of sample collection, yet they are critical to a successful field campaign.
This chapter was written in full awareness that “good practices” are everevolving. The relative importance of, and acquisition methods for, ground data,
including ecological data, depends on the research question, on the project goals, as
well as on study scale, spectroscopic methods and RS data used, budget, time, site
accessibility, and the personnel and their training (Justice and Townshend 1981).
Likewise, spectral processing standards evolve and software goes out of date
A. K. Schweiger