433
(i.e., the scale of an individual plant crown), particularly when appropriate sampling and scaling methods are followed, suggesting that scale-appropriate RS
methods can often resolve trait differences associated with functional or species
diversity (Townsend et al. 2013; Asner et al. 2015). This is particularly relevant
to airborne data, where pixel sizes often approximate those of individual tree
crowns, allowing individual traits to be distinguished (Asner et al. 2015; Singh
et al. 2015), but calls into question the accuracy of trait retrievals when the pixel
sizes are much larger than individual canopies, as is the case for most satellite RS.
Not surprisingly, a review of the recent literature suggests strong effects of
spatial scale on the ability to detect biodiversity using optical RS. Most of this
research has been conducted in North American prairie, with relatively shortstatured vegetation, so may be biased by the relatively small plant crown sizes
(roughly 10 cm). One set of studies, conducted using tallgrass prairie species at
the Cedar Creek Ecosystem Science Reserve in Minnesota (USA), used pixel
sizes ranging from 1 mm (sampled on the ground) to 1 m (sampled from aircraft)
and found a significant correlation between optical diversity and alpha diversity
at finer scales (1 mm to 10 cm), but most of the information on alpha diversity
was lost at pixel sizes larger than about 10 cm, roughly the size of many plant
crowns (Fig. 16.3, red line). Another study conducted at Wood River in Nebraska
(USA) using airborne data found a strong relationship (R
2
) between optical
diversity and alpha diversity for prairie species at pixel sizes of 0.5–1 m and
noted that the optical diversity-alpha diversity relationship was markedly weakened at progressively larger spatial scales (up to 6 m) (Fig. 16.3, black line). A
third study, conducted at Mattheis Ranch in southern Alberta (Canada), found an
intermediate relationship between the two other prairie sites (Fig. 16.3, blue
square).
These studies have significant implications for attempts to sample optical diversity from aircraft or satellite sensors and illustrate the importance of matching sampling scale (pixel size) to crown size when designing airborne campaigns. Clearly,
there is a strong scale-dependence of the optical diversity-biodiversity relationship,
but this scale-dependence varies between study sites, even for the same biome (prairie grassland, in this case). These site differences have been attributed to a number
of possible factors, including the degree of disturbance (fire regimes, invasion by
exotic weeds, or subsequent weed removal), the size of the plot (sampling extent)
relative to the pixel (grain) size, and the differences in species richness between
studies. The Wood River site was less disturbed, with less bare ground and larger,
more species-rich plots than the Cedar Creek site (Gholizadeh et al. 2019). It is
likely that multiple features influence the scale dependence of the optical diversitybiodiversity relationship, illustrating that the larger context and experimental design
of a study can matter.
We know less about the scale dependence of these relationships for other biomes
where explicit scaling experiments involving the RS of biodiversity have not yet
been conducted. Despite unanswered questions, these experiments in prairie ecosystems demonstrate the importance of spatial scale and support the idea that can16 Consideration of Scale in Remote Sensing of Biodiversity
(i.e., the scale of an individual plant crown), particularly when appropriate sampling and scaling methods are followed, suggesting that scale-appropriate RS
methods can often resolve trait differences associated with functional or species
diversity (Townsend et al. 2013; Asner et al. 2015). This is particularly relevant
to airborne data, where pixel sizes often approximate those of individual tree
crowns, allowing individual traits to be distinguished (Asner et al. 2015; Singh
et al. 2015), but calls into question the accuracy of trait retrievals when the pixel
sizes are much larger than individual canopies, as is the case for most satellite RS.
Not surprisingly, a review of the recent literature suggests strong effects of
spatial scale on the ability to detect biodiversity using optical RS. Most of this
research has been conducted in North American prairie, with relatively shortstatured vegetation, so may be biased by the relatively small plant crown sizes
(roughly 10 cm). One set of studies, conducted using tallgrass prairie species at
the Cedar Creek Ecosystem Science Reserve in Minnesota (USA), used pixel
sizes ranging from 1 mm (sampled on the ground) to 1 m (sampled from aircraft)
and found a significant correlation between optical diversity and alpha diversity
at finer scales (1 mm to 10 cm), but most of the information on alpha diversity
was lost at pixel sizes larger than about 10 cm, roughly the size of many plant
crowns (Fig. 16.3, red line). Another study conducted at Wood River in Nebraska
(USA) using airborne data found a strong relationship (R
2
) between optical
diversity and alpha diversity for prairie species at pixel sizes of 0.5–1 m and
noted that the optical diversity-alpha diversity relationship was markedly weakened at progressively larger spatial scales (up to 6 m) (Fig. 16.3, black line). A
third study, conducted at Mattheis Ranch in southern Alberta (Canada), found an
intermediate relationship between the two other prairie sites (Fig. 16.3, blue
square).
These studies have significant implications for attempts to sample optical diversity from aircraft or satellite sensors and illustrate the importance of matching sampling scale (pixel size) to crown size when designing airborne campaigns. Clearly,
there is a strong scale-dependence of the optical diversity-biodiversity relationship,
but this scale-dependence varies between study sites, even for the same biome (prairie grassland, in this case). These site differences have been attributed to a number
of possible factors, including the degree of disturbance (fire regimes, invasion by
exotic weeds, or subsequent weed removal), the size of the plot (sampling extent)
relative to the pixel (grain) size, and the differences in species richness between
studies. The Wood River site was less disturbed, with less bare ground and larger,
more species-rich plots than the Cedar Creek site (Gholizadeh et al. 2019). It is
likely that multiple features influence the scale dependence of the optical diversitybiodiversity relationship, illustrating that the larger context and experimental design
of a study can matter.
We know less about the scale dependence of these relationships for other biomes
where explicit scaling experiments involving the RS of biodiversity have not yet
been conducted. Despite unanswered questions, these experiments in prairie ecosystems demonstrate the importance of spatial scale and support the idea that can16 Consideration of Scale in Remote Sensing of Biodiversity