30
Exploit the Strengths of Remote Sensing
Vegetation types are complex phenomena. Over the course of decades, ecologists
have developed ways of studying and describing this complexity, mostly involving
the classification of a continuous phenomenon into vegetation types (i.e., categories
characterized by plant species and their abundances). Through careful study, temporal changes in vegetation types have also been linked to changes in environmental
conditions and, further, to pressures and driving forces causing these changes. This
provided a framework for monitoring vegetation status, which was eagerly adopted
in Natura 2000 monitoring.
With the advent of remote sensing, attempts were (and are) made to apply this
new technology as an alternative data source for established monitoring methods.
However, this is not necessarily the best way to exploit remote sensing. Vanden
Borre et al. (2011b) already pointed out that remote sensing may be far more suited
for studying vegetation, and unravelling linkages, in ways that were previously
unimaginable. However, this requires active involvement and forward-thinking creativity of both users and method developers to come up with methods that do not
solely replace existing monitoring schemes but instead result in new approaches to
biodiversity monitoring in which remote sensing methods are integrated with other
methods, such as field surveys.
In recent years, a number of such approaches have emerged. For instance, Buck
et al. (2015) identified features of relevance for Natura 2000 grasslands, and mapped
these as raster information layers over larger areas from various remote sensing or
other sources. These information layers constitute a type of primitives that can be
flexibly used by ecologists to infer conclusions about grassland habitat types, such as
probability of occurrence, intensity of use, threats, etc. Since these information layers
(e.g., patch size and shape, spectral homogeneity, line structures, temporal profile of
biomass) are usually derived from established remote sensing methods, they are less
error-prone than a direct grassland habitat classification, and more flexible in accommodating changed relationships (e.g., over space or time) between these information
layers and the habitat types. Hence, they are expected to be more easily transferable.
Lucas et al. (2015) used the principle of data primitives in various stages of their
EODHaM system, a comprehensive method aimed at consistent mapping of land
cover and Natura 2000 habitat types in and around Natura 2000 sites. Whereas
obtaining the data primitives from remote sensing data is more or less straightforward (e.g., spectral indices, but also: size, shape and density of small objects within
larger objects, e.g., tree crowns within a forest versus an orchard), their translation
to Natura 2000 habitat types is achieved through rulesets based on expert input,
which allows for a more flexible adaptation to different geographical settings.
J. Vanden Borre et al.
Exploit the Strengths of Remote Sensing
Vegetation types are complex phenomena. Over the course of decades, ecologists
have developed ways of studying and describing this complexity, mostly involving
the classification of a continuous phenomenon into vegetation types (i.e., categories
characterized by plant species and their abundances). Through careful study, temporal changes in vegetation types have also been linked to changes in environmental
conditions and, further, to pressures and driving forces causing these changes. This
provided a framework for monitoring vegetation status, which was eagerly adopted
in Natura 2000 monitoring.
With the advent of remote sensing, attempts were (and are) made to apply this
new technology as an alternative data source for established monitoring methods.
However, this is not necessarily the best way to exploit remote sensing. Vanden
Borre et al. (2011b) already pointed out that remote sensing may be far more suited
for studying vegetation, and unravelling linkages, in ways that were previously
unimaginable. However, this requires active involvement and forward-thinking creativity of both users and method developers to come up with methods that do not
solely replace existing monitoring schemes but instead result in new approaches to
biodiversity monitoring in which remote sensing methods are integrated with other
methods, such as field surveys.
In recent years, a number of such approaches have emerged. For instance, Buck
et al. (2015) identified features of relevance for Natura 2000 grasslands, and mapped
these as raster information layers over larger areas from various remote sensing or
other sources. These information layers constitute a type of primitives that can be
flexibly used by ecologists to infer conclusions about grassland habitat types, such as
probability of occurrence, intensity of use, threats, etc. Since these information layers
(e.g., patch size and shape, spectral homogeneity, line structures, temporal profile of
biomass) are usually derived from established remote sensing methods, they are less
error-prone than a direct grassland habitat classification, and more flexible in accommodating changed relationships (e.g., over space or time) between these information
layers and the habitat types. Hence, they are expected to be more easily transferable.
Lucas et al. (2015) used the principle of data primitives in various stages of their
EODHaM system, a comprehensive method aimed at consistent mapping of land
cover and Natura 2000 habitat types in and around Natura 2000 sites. Whereas
obtaining the data primitives from remote sensing data is more or less straightforward (e.g., spectral indices, but also: size, shape and density of small objects within
larger objects, e.g., tree crowns within a forest versus an orchard), their translation
to Natura 2000 habitat types is achieved through rulesets based on expert input,
which allows for a more flexible adaptation to different geographical settings.
J. Vanden Borre et al.