60
(Falster and Westoby 2003), with consequences for interpreting optical RS signatures (Ollinger 2011). Thus, when considering the use of RS approaches for mapping leaf traits, careful consideration of vegetation structure, collection
characteristics, and sensor design is important.
Phenology, leaf seasonality, and leaf age are also important drivers of optical
properties for a number of reasons. First, leaf traits can change significantly over the
lifetime of a leaf (e.g., Wilson et al. 2001; Niinemets 2016; Chavana-Bryant et al.
2017; Wu et al. 2017), and the corresponding leaf optical properties will change in
concert (Yang et al. 2016). Average leaf angle distribution can also change with leaf
age or seasonally from younger, recently expanded leaves to fully expanded (Raabe
et al. 2015), which can have significant impacts on canopy reflectance (Huemmrich
2013). Finally, atmospheric, insect, or other stressors typically change the chemical
makeup of leaves and so their optical properties (e.g., Couture et al. 2013; Ainsworth
et al. 2014; Cotrozzi et al. 2018).
3.2.2 Approaches for Linking Traits and Spectral Signatures
Despite the promise and utility of spectroscopy for the retrieval and mapping of
plant traits across space and time, there has not been consensus or standardization
of approaches and algorithm development in the RS and biodiversity communities.
This is not entirely unexpected given the complexity of connecting traits and RS
observations across the various scales of interest, from leaves to individual trees,
communities, and landscapes (Schweiger, Chap. 15). In addition, early approaches
(e.g., Peterson et al. 1988) were often later deemed inappropriate and often replaced
by other techniques (e.g., Grossman et al. 1996). Access to more powerful, improved,
and cheaper computing resources has also allowed for the exploration of more complex statistical and machine-learning approaches (see Schweiger, Chap. 15).
Two primary approaches have been utilized to link RS observations to functional
traits—empirical, statistically based techniques and radiative transfer modeling
(RTM; see also Meireles et al., Chap. 7; Ustin, Chap. 14).
3.2.2.1 Empirical Scaling Approaches
With respect to empirical techniques, the use of SVIs was one of the earliest methods
to explore the capacity to link a range of plant functional traits to vegetation spectra.
Typically, with this approach a single SVI is linked with a trait of interest, such as
leaf pigments or water content, to develop a simple statistical relationship between the
trait of interest and corresponding variation in optical properties (e.g., Sims and
Gamon 2003; Gitelson 2004; Colombo et al. 2008; Feret et al. 2011). The derived
model is then used to estimate trait values for new leaves using only spectral measurements. This approach typically assumes the researcher has an a priori understanding of the links between the trait and resulting variation in the electromagnetic
S. P. Serbin and P. A. Townsend
(Falster and Westoby 2003), with consequences for interpreting optical RS signatures (Ollinger 2011). Thus, when considering the use of RS approaches for mapping leaf traits, careful consideration of vegetation structure, collection
characteristics, and sensor design is important.
Phenology, leaf seasonality, and leaf age are also important drivers of optical
properties for a number of reasons. First, leaf traits can change significantly over the
lifetime of a leaf (e.g., Wilson et al. 2001; Niinemets 2016; Chavana-Bryant et al.
2017; Wu et al. 2017), and the corresponding leaf optical properties will change in
concert (Yang et al. 2016). Average leaf angle distribution can also change with leaf
age or seasonally from younger, recently expanded leaves to fully expanded (Raabe
et al. 2015), which can have significant impacts on canopy reflectance (Huemmrich
2013). Finally, atmospheric, insect, or other stressors typically change the chemical
makeup of leaves and so their optical properties (e.g., Couture et al. 2013; Ainsworth
et al. 2014; Cotrozzi et al. 2018).
3.2.2 Approaches for Linking Traits and Spectral Signatures
Despite the promise and utility of spectroscopy for the retrieval and mapping of
plant traits across space and time, there has not been consensus or standardization
of approaches and algorithm development in the RS and biodiversity communities.
This is not entirely unexpected given the complexity of connecting traits and RS
observations across the various scales of interest, from leaves to individual trees,
communities, and landscapes (Schweiger, Chap. 15). In addition, early approaches
(e.g., Peterson et al. 1988) were often later deemed inappropriate and often replaced
by other techniques (e.g., Grossman et al. 1996). Access to more powerful, improved,
and cheaper computing resources has also allowed for the exploration of more complex statistical and machine-learning approaches (see Schweiger, Chap. 15).
Two primary approaches have been utilized to link RS observations to functional
traits—empirical, statistically based techniques and radiative transfer modeling
(RTM; see also Meireles et al., Chap. 7; Ustin, Chap. 14).
3.2.2.1 Empirical Scaling Approaches
With respect to empirical techniques, the use of SVIs was one of the earliest methods
to explore the capacity to link a range of plant functional traits to vegetation spectra.
Typically, with this approach a single SVI is linked with a trait of interest, such as
leaf pigments or water content, to develop a simple statistical relationship between the
trait of interest and corresponding variation in optical properties (e.g., Sims and
Gamon 2003; Gitelson 2004; Colombo et al. 2008; Feret et al. 2011). The derived
model is then used to estimate trait values for new leaves using only spectral measurements. This approach typically assumes the researcher has an a priori understanding of the links between the trait and resulting variation in the electromagnetic
S. P. Serbin and P. A. Townsend