418
of overflight, such that each flight line can be corrected separately. However, when
multiple sets of tarps are not an option, a single set of tarps can be placed in a central
area instead. Other surfaces can be used for ELC, including water bodies, road intersections, rooftops, snow, and ice. Ideally, these targets are distributed throughout the
image and are measured at the time of the overpass. However, it might be difficult
to find already existing calibration targets that are spectrally stable and uniform; for
example, the reflectance of lakes can vary substantially depending on the distance
from the shore and currents; snow and ice are often less spectrally uniform than
expected due to surface irregularities, dust, and algae growth; and asphalt concrete
varies spectrally depending on the aggregate composition (bitumen varies less). If
possible, limiting movement and traffic around calibration targets during measurements and image acquisition is advantageous for reducing the amount of dust and
dirt (e.g., when using calibration tarps or roads), as well as limiting surface disturbance and shadow cast (e.g., when using snow). However, prioritizing calibration
target quality over quantity is good practice. For example, measuring one set of
spectrally stable tarps in a flat area in the center of an image is preferable over measuring several natural calibration targets that are likely not as spectrally uniform and
might be difficult to locate precisely.
15.4 Conclusions
As ecologists and remote sensing scientists are coming together to develop methods
for the continuous assessment and monitoring of plant biodiversity, connecting the
local to the global scale, studies of species to communities and ecosystems, and
ecological resources to human needs and values become more and more feasible.
Field campaigns are essential to this effect, because only (the repeated process of
generating) ecological knowledge and data (including spectral measurements of
plants) make it possible to understand better what is happening to the ecosystems
and species we care about and why. Thanks to the ecological data revolution, remote
sensing and organismal data as well as climate, land use, and socioeconomic data
are becoming increasingly publicly available. At the same time, bioinformatics and
cyberinfrastructure promote innovative ways for data handling, storage, and visualization and for integrating and analyzing these data across scales. Harnessing this
amount of data requires developing and documenting data standards that facilitate
collaborations across disciplines, data integration across sites and scales, data discovery for meta-analysis, and model re-calibrations. This makes the planning of
data collection with consideration of the data life cycle as important as the data collection itself. It is important in the face of our current and future challenges and
opens a wealth of opportunities in biodiversity science.
Acknowledgments AKS wishes to thank Susan L. Ustin and Shawn P. Serbin for reviewing this
chapter and the National Institute for Mathematical Biology and Synthesis (NIMBioS) working
group on remote sensing for stimulating discussions. This work was supported by the NSF/NASA
A. K. Schweiger
of overflight, such that each flight line can be corrected separately. However, when
multiple sets of tarps are not an option, a single set of tarps can be placed in a central
area instead. Other surfaces can be used for ELC, including water bodies, road intersections, rooftops, snow, and ice. Ideally, these targets are distributed throughout the
image and are measured at the time of the overpass. However, it might be difficult
to find already existing calibration targets that are spectrally stable and uniform; for
example, the reflectance of lakes can vary substantially depending on the distance
from the shore and currents; snow and ice are often less spectrally uniform than
expected due to surface irregularities, dust, and algae growth; and asphalt concrete
varies spectrally depending on the aggregate composition (bitumen varies less). If
possible, limiting movement and traffic around calibration targets during measurements and image acquisition is advantageous for reducing the amount of dust and
dirt (e.g., when using calibration tarps or roads), as well as limiting surface disturbance and shadow cast (e.g., when using snow). However, prioritizing calibration
target quality over quantity is good practice. For example, measuring one set of
spectrally stable tarps in a flat area in the center of an image is preferable over measuring several natural calibration targets that are likely not as spectrally uniform and
might be difficult to locate precisely.
15.4 Conclusions
As ecologists and remote sensing scientists are coming together to develop methods
for the continuous assessment and monitoring of plant biodiversity, connecting the
local to the global scale, studies of species to communities and ecosystems, and
ecological resources to human needs and values become more and more feasible.
Field campaigns are essential to this effect, because only (the repeated process of
generating) ecological knowledge and data (including spectral measurements of
plants) make it possible to understand better what is happening to the ecosystems
and species we care about and why. Thanks to the ecological data revolution, remote
sensing and organismal data as well as climate, land use, and socioeconomic data
are becoming increasingly publicly available. At the same time, bioinformatics and
cyberinfrastructure promote innovative ways for data handling, storage, and visualization and for integrating and analyzing these data across scales. Harnessing this
amount of data requires developing and documenting data standards that facilitate
collaborations across disciplines, data integration across sites and scales, data discovery for meta-analysis, and model re-calibrations. This makes the planning of
data collection with consideration of the data life cycle as important as the data collection itself. It is important in the face of our current and future challenges and
opens a wealth of opportunities in biodiversity science.
Acknowledgments AKS wishes to thank Susan L. Ustin and Shawn P. Serbin for reviewing this
chapter and the National Institute for Mathematical Biology and Synthesis (NIMBioS) working
group on remote sensing for stimulating discussions. This work was supported by the NSF/NASA
A. K. Schweiger