393
characteristics. In “forward mode,” RTMs simulate spectra from the vegetation
parameters incorporated into the model, e.g., by using the expected range of values
for chlorophyll content, SLA, and other parameters, in certain ecosystems or for
certain species. In “backward mode,” RTMs estimate vegetation characteristics
from spectra (Vohland and Jarmer 2008; Weiss et al. 2000). Model inversion is usually done by running an RTM in forward mode and systematically varying the input
parameters using lookup tables (LUT) with different trait combinations, until the
measured spectral signal is sufficiently well approximated. The most plausible input
parameter combinations can then be averaged to provide estimates of vegetation
traits. In principle, RTM inversion can be conducted without ground reference data
collected on-site (e.g., when input parameters can be sourced from plant trait databases, such as TRY; Kattge et al. 2011) or determined based on expert knowledge.
However, inversion of RTMs is generally an ill-posed problem in the sense that
there is not a single solution but rather multiple solutions to model inversions (i.e.,
multiple input parameters can yield the same output spectra). Ground reference data
are important for RTMs because they can be used to limit the ranges of possible
input values (Combal et al. 2003) and are essential for model validation.
15.2 Planning Field Campaigns
This section includes thoughts about data organization (Sect. 15.2.1) and communication (Sect. 15.2.2), before covering the planning of data collection in more detail
(Sect. 15.2.3).
15.2.1 Data Organization
Data organization schemes help define and implement guidelines to make project
management and collaborations more efficient and ensure long-term project sustainability and the reproducibility of research (https://ropensci.github.io/reproducibility-guide/). Guidelines for folder structure, file names, documentation, file
formats, data sharing and archiving, version control, and data backups are all part of
data organization. When archiving is handled by a third party, researchers need to
consider how to structure data and metadata to match external requirements.
Generally, it is good practice to work backward and start with identifying where
project data will be stored long term and which data and metadata standards will
make long-term storage possible and data sets discoverable and reusable later. For
instance, it is important to use community standards for taxon names, units, and
keywords and to store data in file formats that are nonproprietary (open), unencrypted, and in common use by the research community. The US Library of
Congress has released a recommended file format statement (http://www.loc.gov/
15 Spectral Field Campaigns: Planning and Data Collection
characteristics. In “forward mode,” RTMs simulate spectra from the vegetation
parameters incorporated into the model, e.g., by using the expected range of values
for chlorophyll content, SLA, and other parameters, in certain ecosystems or for
certain species. In “backward mode,” RTMs estimate vegetation characteristics
from spectra (Vohland and Jarmer 2008; Weiss et al. 2000). Model inversion is usually done by running an RTM in forward mode and systematically varying the input
parameters using lookup tables (LUT) with different trait combinations, until the
measured spectral signal is sufficiently well approximated. The most plausible input
parameter combinations can then be averaged to provide estimates of vegetation
traits. In principle, RTM inversion can be conducted without ground reference data
collected on-site (e.g., when input parameters can be sourced from plant trait databases, such as TRY; Kattge et al. 2011) or determined based on expert knowledge.
However, inversion of RTMs is generally an ill-posed problem in the sense that
there is not a single solution but rather multiple solutions to model inversions (i.e.,
multiple input parameters can yield the same output spectra). Ground reference data
are important for RTMs because they can be used to limit the ranges of possible
input values (Combal et al. 2003) and are essential for model validation.
15.2 Planning Field Campaigns
This section includes thoughts about data organization (Sect. 15.2.1) and communication (Sect. 15.2.2), before covering the planning of data collection in more detail
(Sect. 15.2.3).
15.2.1 Data Organization
Data organization schemes help define and implement guidelines to make project
management and collaborations more efficient and ensure long-term project sustainability and the reproducibility of research (https://ropensci.github.io/reproducibility-guide/). Guidelines for folder structure, file names, documentation, file
formats, data sharing and archiving, version control, and data backups are all part of
data organization. When archiving is handled by a third party, researchers need to
consider how to structure data and metadata to match external requirements.
Generally, it is good practice to work backward and start with identifying where
project data will be stored long term and which data and metadata standards will
make long-term storage possible and data sets discoverable and reusable later. For
instance, it is important to use community standards for taxon names, units, and
keywords and to store data in file formats that are nonproprietary (open), unencrypted, and in common use by the research community. The US Library of
Congress has released a recommended file format statement (http://www.loc.gov/
15 Spectral Field Campaigns: Planning and Data Collection