Environmental Engineering Reference
In-Depth Information
even when high-resolution remote sensor data is available, it
is difficult to determine species richness without conducting
field surveys. In order to study biodiversity in urban areas (and,
indeed, elsewhere) cost efficient combinations of the two meth-
ods are required; this has been the case in most previous studies
of biodiversity using remote sensing (see Gottschalk, Huettmann
and Ehlers, 2005).
Anothermain advantage of using remote sensing for biodiver-
sity studies is the possibility of providing data for spatial-temporal
analyses of landscapes. Aerial photographs are stored in many
national archives and date from at least the early 1940s, while
imaging from space has been widely collected since the 1970s.
Furthermore, within the temporal domain provided by many
satellite sensors, with repeat periods of between 15 minutes and
a few weeks, it is also possible to undertake biodiversity studies
of the monthly, seasonal and annual dynamics of landscapes
(Groom et al ., 2006).
Capturing biodiversity information through remote sensor
can be divided into direct and indirect approaches (Turner
et al ., 2003, Duro et al ., 2007). According to these authors,
direct approaches are first-order analyses of the occurrence
of species or species assemblages using remote sensing, while
indirect approaches use remote sensor data to measure
environmental variables that are related to biodiversity. The
research into the two approaches is reviewed in the following
subsections.
Since the late 1990s, there has been a major increase in
the availability of digital satellite imagery, with very high spa-
tial resolution, i.e., less than 5 m. Satellites that provide such
high resolution data include Quickbird (0.6-2.5 m), IKONOS
(1-4 m), and SPOT (2.5 m). There is great potential for man-
ually or digitally identifying tree species and canopy attributes
from these sources (e.g., Clark et al ., 2004; Levin et al ., 2009).
In a comparison between manual and automated interpretation
of CIR aerial images and IKONOS satellite images, the manual
interpretation of CIR aerial images delivered data with the highest
accuracy in classes with a very high information value for bio-
diversity studies (Groom et al ., 2006). However, with the lower
cost and greater availability of high-resolution satellite images
and the rapid development of new classification techniques, in
combination with other data sources, their value for biodiver-
sity studies can be expected to increase significantly. Promising
research on new classification techniques embrace, for example,
computerized approaches and the use of multilayer perception
and neural networks that significantly improve accuracy (e.g.,
Boyd, Sanchez-Hernandez and Foody, 2006; Sesnie et al ., 2008)
and object-based image segmentation (e.g., Burnett andBlaschke,
2003), thus moving away from the pixel-based approach.
Newdevelopments in retrieving biodiversity-related informa-
tion from remote sensor data are facilitated by major improve-
ments in data capture. Data from active sensors, such as radar,
can provide information on elevation, e.g., from the Shuttle
Radar Topography Mission which has almost global coverage;
these data are valuable for biodiversity studies. Radar can also
provide high-resolution data, for example, Saatchi et al . (2008)
used radar backscatter from QSCAT to improve models by
providing information on vegetation structure. Taft, Haig and
Kiilsgaard (2003) and Lang et al . (2008) used radar remote sensor
data for wetland mapping, with resolutions of 5.6 to 68 cm in
the bands used. Another type of active sensor is airborne lidar,
which has been used to improve species distribution models by
quantifying vegetation structure within a landscape (Hill and
Thompson, 2005; Goetz et al ., 2007) and for tree species classifi-
cation with a resolution of 0.2 to 2 m (Hyyppa et al ., 2008). Other
advances are associated with techniques such as multi-angle
viewing (e.g., Baltsavias et al ., 2008) and hyperspectral remote
sensing (e.g., Ben-Dor, Levin and Saaroni, 2001; Foody et al .,
2004) that have considerable potential relevant to biodiversity
studies. Finally, combinations of data from different sensors can
be combined to enhance the information derived from them, e.g.
laser scanning and multispectral remote sensing (e.g., Holmgren
et al ., 2008).
20.2.1 Direct approaches
Directly capturing information about biodiversity components
fromremote sensing is possible at a range of scales, fromrelatively
coarse classifications of species assemblages or vegetation types
to identifying single species at the scale of individual tree crowns
(e.g., Foody et al ., 2005). Land cover classification, although
often used for deriving biodiversity-related information, can
be considered to be more closely connected to the indirect
methods, and therefore is considered under that heading. Direct
methods tend to require high resolution imagery, from airplanes
or satellites.
Aerial photographs have been collected since at least the early
1940s in many countries. Recently, the availability and quality
of digital image data produced from aerial photographs has
increased, often with national coverage and at resolutions of
less than 1 m (Groom et al ., 2006). Using color infrared (CIR)
aerial photographs, methods have been developed for manual
interpretation in order to obtain detailed vegetation maps with
ecologically meaningful classifications, resulting in data with a
very high information value for biodiversity studies (Ihse, 1995).
These methods were further developed by Lofvenhaft, Bjorn and
Ihse (2002), with the specific aim of supporting biodiversity
issues relating to spatial planning in urban and suburban areas
in Stockholm, Sweden. Apart from retrieving data pertaining
to land cover and detailed vegetation types, they also captured
data on habitat structures and vegetation cover within built-up
areas. The accuracy compared to the field control was very high
(93-95%) for the main land cover types and for vegetation type
classes; for comparison, classes relating to habitat structure in
hardwood deciduous forest were identified at an accuracy of
72-75%. This database has proved to be very valuable for the use
in urban planning (Mortberg and Ihse, 2006).
20.2.2 Indirect approaches
Despite the advances in extracting biodiversity-relevant infor-
mation directly from remote sensor data, given the extremely
wide definition of biodiversity, we can only expect a fraction of
all biodiversity components to be detected directly by remote
sensing. However, remote sensor data provide an invaluable and
yet underused information source relating to biodiversity and
ecological processes, which could be accessed through indirect
approaches including spatial ecological modeling (e.g., Turner
et al ., 2003). In this way, it is possible to explore relation-
ships between biodiversity components, ecological processes and
predictive environmental variables derived from remote sensing
Search WWH ::




Custom Search