28
mized when there are more units in a community that are more equitably distributed
(or less clumped) and more dispersed (or positioned further apart) in space. Like
Chao’s approach,
q
D(TM) includes Hill numbers (
q
), which allow weighting of
abundances: small and large q values emphasize rare and common species, respectively. Like many other biodiversity metrics,
q
D(TM) can be calculated from pairwise distances among species or individuals; thus, the metric can be applied to
estimate different dimensions of biodiversity, including functional, phylogenetic
(Scheiner 2012; Presley et al. 2014) and spectral components (Schweiger et al. 2018).
Briefly, functional trait dispersion [
q
D(TM)] is calculated as:
q
q
D TM
S
E T M
u u
1
1
’
(2.4)
where:
S = species richness
E(T) = trait evenness
M’ = trait dispersion
q = Hill number
2.7.5 Spectral Diversity
Like taxonomic, functional, and phylogenetic diversity, spectral diversity can be
calculated in many different ways. Spectral alpha diversity metrics include the
coefficient of variation of spectral indices (Oindo and Skidmore 2002) or spectral
bands among pixels (Hall et al. 2010; Gholizadeh et al. 2018, 2019; Wang et al.
2018, the convex hull volume (Dahlin 2016) and the convex hull area (Gholizadeh
et al. 2018) of pixels in spectral feature space, the mean distance of pixels from the
spectral centroid (Rocchini et al. 2010), the number of spectrally distinct clusters
or “spectral species” in ordination space (Féret and Asner 2014), and spectral variance (Laliberté et al. 2019). Schweiger et al. (2018) applied
q
D(TM) to species
mean spectra and to individual pixels extracted at random from high-resolution
proximal RS data. The second approach is independent of species identity and uses
the same number of pixels per community for analysis. In this manner, the problem
of diversity scaling with the number of species in a community is eliminated, and
greater differences in reflectance spectra among pixels result in increased spectral
diversity. Conceptually, spectral diversity metrics are versatile and can be tailored
to match taxonomic or phylogenetic units, e.g., by using mean spectra for focal
taxa, or to resemble functional diversity by selecting spectral bands that align with
known absorption features for specific chemical traits or spectral indices that capture plant characteristics of known ecological importance. If measured at the
appropriate scale (see Gamon et al. Chap. 16), spectral diversity can integrate the
variation captured by other metrics of diversity and similarly predicts ecosystem
function (Fig. 2.6).
J. Cavender-Bares et al.
mized when there are more units in a community that are more equitably distributed
(or less clumped) and more dispersed (or positioned further apart) in space. Like
Chao’s approach,
q
D(TM) includes Hill numbers (
q
), which allow weighting of
abundances: small and large q values emphasize rare and common species, respectively. Like many other biodiversity metrics,
q
D(TM) can be calculated from pairwise distances among species or individuals; thus, the metric can be applied to
estimate different dimensions of biodiversity, including functional, phylogenetic
(Scheiner 2012; Presley et al. 2014) and spectral components (Schweiger et al. 2018).
Briefly, functional trait dispersion [
q
D(TM)] is calculated as:
q
q
D TM
S
E T M
u u
1
1
’
(2.4)
where:
S = species richness
E(T) = trait evenness
M’ = trait dispersion
q = Hill number
2.7.5 Spectral Diversity
Like taxonomic, functional, and phylogenetic diversity, spectral diversity can be
calculated in many different ways. Spectral alpha diversity metrics include the
coefficient of variation of spectral indices (Oindo and Skidmore 2002) or spectral
bands among pixels (Hall et al. 2010; Gholizadeh et al. 2018, 2019; Wang et al.
2018, the convex hull volume (Dahlin 2016) and the convex hull area (Gholizadeh
et al. 2018) of pixels in spectral feature space, the mean distance of pixels from the
spectral centroid (Rocchini et al. 2010), the number of spectrally distinct clusters
or “spectral species” in ordination space (Féret and Asner 2014), and spectral variance (Laliberté et al. 2019). Schweiger et al. (2018) applied
q
D(TM) to species
mean spectra and to individual pixels extracted at random from high-resolution
proximal RS data. The second approach is independent of species identity and uses
the same number of pixels per community for analysis. In this manner, the problem
of diversity scaling with the number of species in a community is eliminated, and
greater differences in reflectance spectra among pixels result in increased spectral
diversity. Conceptually, spectral diversity metrics are versatile and can be tailored
to match taxonomic or phylogenetic units, e.g., by using mean spectra for focal
taxa, or to resemble functional diversity by selecting spectral bands that align with
known absorption features for specific chemical traits or spectral indices that capture plant characteristics of known ecological importance. If measured at the
appropriate scale (see Gamon et al. Chap. 16), spectral diversity can integrate the
variation captured by other metrics of diversity and similarly predicts ecosystem
function (Fig. 2.6).
J. Cavender-Bares et al.