practice includes evaluating the effect of any assumptions underlying source data on the
results of the synthesis product.
Data synthesis can be more complex than combining data layers in a GIS model.
New software tools for object oriented classification, such as eCognition, increase the
accessibility of 2
nd generation remote sensing techniques where images are divided into
uniform regions at various scales prior to classification (Benz et al., 2004). Applications
of object-oriented approaches to coastal management are now being emphasized,
improving basic remote sensing classifications in local-scale studies. In one study,
Kaya et al. (2002) used object-oriented classifications of multitemporal Radarsat-1 data
to classify populated areas and wetlands, and then used the data in a GIS analysis of
malaria risk in coastal Kenya. Beyond multitemporal data, object-oriented approaches
can also be used to combine several datasets. For example, data layers from one map
product could be used to divide a remotely sensed image into segments for subsequent
classifications. The ability to use segmentation to separately evaluate areas with
different inherent spatial scales will allow new approaches to data synthesis including
remote sensing data (Schiewe et al., 2001).
5. Data Distribution and Impediments to Distribution
A key element in the global and regional sources of data discussed in this chapter is
accessibility. Both technical and social issues can influence the accessibility of data for
synthesis. Technical issues include issues of data volume, security, format, projection,
Internet map server (IMS) technologies, and accessibility and quality of metadata.
Social issues relate to data ownership, credit, and stake in synthesis products.
For regional and global studies, the size of datasets can be an issue impeding the
distribution and use of data, especially for high spatial resolution datasets. Errors in
data conversions or poor methodological choices can also increase the size of datasets,
slowing subsequent analyses. For example, we have encountered enormous ESRI
Shapefiles (Environmental Systems Research Institute, Inc.) where every pixel in the
raster data had been converted into an individual polygon. The size of datasets is a
particularly important issue for distribution of data online in IMS applications
(including Arc/IMS, Demis Map Server, and open source GIS). IMS technology is
often slow even in areas with high Internet bandwidth—but slow data transmission
speeds can make online datasets completely inaccessible in areas with only dial-up
access to the Internet (areas which may be most in need of access to shared data).
Large global datasets also can have significant costs for data storage and distribution
that can impact the ability to distribute data at no cost over the long term. Products
distributed via DVDs or CDs also have cumulative costs of distribution.
Data format and projection can also be issues. Both proprietary and
collaborative formats have key roles in the exchange of map information, with
ESRI (1998) Shapefiles leading for vector data and open GeoTIFF standards
(http://remotesensing.org/geotiff/geotiff.html) for raster data.
A proprietary
compression standard, MrSID (Multi-resolution Seamless Image Database) allows
selective decompression and use of extremely large remote sensing images over
networks. At present, a number of software packages can view MrSID files, but
licensing to create the files is priced based on the amount of data encoded
(http://www.lizardtech.com). MrSID has been used as the distribution format for
hich are of benefit as base remote sensing information for hundreds of applications.
orthorectified Landsat mosaics produced by Earthsat for NASA (Tucker et al.,2004),
w
297
Data Synthesis for Management
results of the synthesis product.
Data synthesis can be more complex than combining data layers in a GIS model.
New software tools for object oriented classification, such as eCognition, increase the
accessibility of 2
nd generation remote sensing techniques where images are divided into
uniform regions at various scales prior to classification (Benz et al., 2004). Applications
of object-oriented approaches to coastal management are now being emphasized,
improving basic remote sensing classifications in local-scale studies. In one study,
Kaya et al. (2002) used object-oriented classifications of multitemporal Radarsat-1 data
to classify populated areas and wetlands, and then used the data in a GIS analysis of
malaria risk in coastal Kenya. Beyond multitemporal data, object-oriented approaches
can also be used to combine several datasets. For example, data layers from one map
product could be used to divide a remotely sensed image into segments for subsequent
classifications. The ability to use segmentation to separately evaluate areas with
different inherent spatial scales will allow new approaches to data synthesis including
remote sensing data (Schiewe et al., 2001).
5. Data Distribution and Impediments to Distribution
A key element in the global and regional sources of data discussed in this chapter is
accessibility. Both technical and social issues can influence the accessibility of data for
synthesis. Technical issues include issues of data volume, security, format, projection,
Internet map server (IMS) technologies, and accessibility and quality of metadata.
Social issues relate to data ownership, credit, and stake in synthesis products.
For regional and global studies, the size of datasets can be an issue impeding the
distribution and use of data, especially for high spatial resolution datasets. Errors in
data conversions or poor methodological choices can also increase the size of datasets,
slowing subsequent analyses. For example, we have encountered enormous ESRI
Shapefiles (Environmental Systems Research Institute, Inc.) where every pixel in the
raster data had been converted into an individual polygon. The size of datasets is a
particularly important issue for distribution of data online in IMS applications
(including Arc/IMS, Demis Map Server, and open source GIS). IMS technology is
often slow even in areas with high Internet bandwidth—but slow data transmission
speeds can make online datasets completely inaccessible in areas with only dial-up
access to the Internet (areas which may be most in need of access to shared data).
Large global datasets also can have significant costs for data storage and distribution
that can impact the ability to distribute data at no cost over the long term. Products
distributed via DVDs or CDs also have cumulative costs of distribution.
Data format and projection can also be issues. Both proprietary and
collaborative formats have key roles in the exchange of map information, with
ESRI (1998) Shapefiles leading for vector data and open GeoTIFF standards
(http://remotesensing.org/geotiff/geotiff.html) for raster data.
A proprietary
compression standard, MrSID (Multi-resolution Seamless Image Database) allows
selective decompression and use of extremely large remote sensing images over
networks. At present, a number of software packages can view MrSID files, but
licensing to create the files is priced based on the amount of data encoded
(http://www.lizardtech.com). MrSID has been used as the distribution format for
hich are of benefit as base remote sensing information for hundreds of applications.
orthorectified Landsat mosaics produced by Earthsat for NASA (Tucker et al.,2004),
w
297
Data Synthesis for Management