399
requires researchers to familiarize themselves with the conditions at the site. It is
also critical to define at an early planning stage how the data will be analyzed and
to choose the sampling design and sample size accordingly (see Sect. 15.2.3.3).
Improper randomization in particular can lead to biased conclusions based on inappropriate assumptions (De Gruijter 1999).
15.2.3.1 Area Selection
It is rarely possible to describe all aspects of an ecological phenomenon of interest
in a single study. Model boundaries help clarify the conditions under which a model
is likely valid (see Sect. 15.1.2). Research areas represent conceptual model boundaries, including ecosystem(s), plant community(ies), species, and geographic location. Although it is good practice to formulate model boundaries first and pick
research areas accordingly, in practice model boundaries usually need to be adjusted
after selecting a research area to reflect the conditions at the site. Formulating the
model first and refining it during the planning process help clarify a study’s limitations; stating them clearly is important for the analysis and further synthesis work.
Once model boundaries have been formulated, it is important to investigate
which environmental conditions influence the response and explanatory variables in
the study system and how they are spatially distributed. Generally, it is critical to
cover the range of environmental conditions, both biotic and abiotic, for which
model inferences are being made, including the diversity and distribution of vegetation communities, plant species, successional gradients, soil types, soil moisture
and nutrient gradients, aspects, slopes, land uses, microclimatic conditions, animal
communities, pathogens, and other factors determining environmental heterogeneity in a study area. Accounting for the variation of every factor might not be possible, but many environmental factors are correlated. If possible, it is good practice to
investigate the covariance structure of environmental factors based on previous
studies and existing data and to focus on a few factors that are expected to have the
most effect on the phenomenon of interest. Data collection might need to be limited
to a smaller area than anticipated due to a high degree of environmental heterogeneity and/or time constraints, which affects the range of conditions for which conclusions can be drawn. However, it is generally advantageous to work with sound
models for small areas with a limited degree of environmental variation than to
work with weaker models for larger areas. Testing model predictions outside the
model boundaries can provide important insights regarding model transferability,
the comparability of ecological conditions, and differences and similarities in ecosystem function between areas.
Maps, local sources, and other research groups can provide important information regarding environmental variation in a research area. Again, it is helpful to
contact site administrators early to gain access to resources and build connections to
other research groups. Local administrators can often give advice regarding the timing and location of sample collection, including practical considerations such as
accessibility. Visiting a potential research area can be extremely helpful during the
15 Spectral Field Campaigns: Planning and Data Collection
requires researchers to familiarize themselves with the conditions at the site. It is
also critical to define at an early planning stage how the data will be analyzed and
to choose the sampling design and sample size accordingly (see Sect. 15.2.3.3).
Improper randomization in particular can lead to biased conclusions based on inappropriate assumptions (De Gruijter 1999).
15.2.3.1 Area Selection
It is rarely possible to describe all aspects of an ecological phenomenon of interest
in a single study. Model boundaries help clarify the conditions under which a model
is likely valid (see Sect. 15.1.2). Research areas represent conceptual model boundaries, including ecosystem(s), plant community(ies), species, and geographic location. Although it is good practice to formulate model boundaries first and pick
research areas accordingly, in practice model boundaries usually need to be adjusted
after selecting a research area to reflect the conditions at the site. Formulating the
model first and refining it during the planning process help clarify a study’s limitations; stating them clearly is important for the analysis and further synthesis work.
Once model boundaries have been formulated, it is important to investigate
which environmental conditions influence the response and explanatory variables in
the study system and how they are spatially distributed. Generally, it is critical to
cover the range of environmental conditions, both biotic and abiotic, for which
model inferences are being made, including the diversity and distribution of vegetation communities, plant species, successional gradients, soil types, soil moisture
and nutrient gradients, aspects, slopes, land uses, microclimatic conditions, animal
communities, pathogens, and other factors determining environmental heterogeneity in a study area. Accounting for the variation of every factor might not be possible, but many environmental factors are correlated. If possible, it is good practice to
investigate the covariance structure of environmental factors based on previous
studies and existing data and to focus on a few factors that are expected to have the
most effect on the phenomenon of interest. Data collection might need to be limited
to a smaller area than anticipated due to a high degree of environmental heterogeneity and/or time constraints, which affects the range of conditions for which conclusions can be drawn. However, it is generally advantageous to work with sound
models for small areas with a limited degree of environmental variation than to
work with weaker models for larger areas. Testing model predictions outside the
model boundaries can provide important insights regarding model transferability,
the comparability of ecological conditions, and differences and similarities in ecosystem function between areas.
Maps, local sources, and other research groups can provide important information regarding environmental variation in a research area. Again, it is helpful to
contact site administrators early to gain access to resources and build connections to
other research groups. Local administrators can often give advice regarding the timing and location of sample collection, including practical considerations such as
accessibility. Visiting a potential research area can be extremely helpful during the
15 Spectral Field Campaigns: Planning and Data Collection