154
Geographic Infonnation Science and Ecological Assessment
TABLE 11.1. Geospatial data types and sources needed for developing vegetation data for ecosystem assessment.
Theme
Scale,
description
Source
Use
Processing
and analysis
Landsat Thematic
Mapper imagery
30 m
NASA, EROS
Data Center (via
Forest Service)
Identify
vegetation
formations and
land cover
Unsupervised
classification,
segmentation
Resource air
photos
Variable;
Forest Service
Conduct field
reconnaissance
and label
vegetation map
features
Aerial photo
interpretation
1 : 12,000 to
I : 30,000
Digital elevation
models (DEMs)
30-m, 7.5'
quadrangles
USGS (via
Forest Service)
Model potential
vegetation
series
Mosaic, smooth,
derive slope, aspect
Cartographic
Feature Files
(CFFs): land
ownership, forest
boundaries,
hydrology, roads
Variable; usually
1: 24,000
Forest Service
(from USGS
digital line
graphs and
various digitized
sources)
Isolate study
area, develop
draft hard-copy
maps for field
checking
Overlay,
georeference, edit
Existing digital
thematic maps
(forest plantations,
previous vegetation
maps)
Variable; usually
1: 24,000
Forest Service
Label certain
features in
vegetation map
(plantations)
Overlay, map editing
Example from a recently completed project to map existing vegetation on Forest Service lands in Southern California carried out
at San Diego State University (Franklin and Woodcock, 1997; Franklin et al., 2000).
tor form as digital line graphs (DLGs). These represent fundamental data for an EA involving any
type of cartographic modeling (discussed later) for
which information such as the distribution of a resource by land ownership or in proximity to roads
is needed. However, as with the raster DEMs discussed previously, these vector data are only as
good as the maps from which they were derived
(which may be outdated and contain errors) and the
quality of the digital conversion process (never free
from errors). For example, in our work in southern
California, we found that the stream network represented in the DLG contained some errors in the
location of streams when compared to georeferenced satellite imagery, but, more importantly,
lacked information on lower-order (intermittent,
seasonal) streams associated with important riparian ecosystems (Franklin et aI., 2000). Owing to
the importance of accurate digital maps of such basic features as land ownership and land use (wilderness, for example) for land management planning
and EA, many federal, state, and regional agencies
are beginning to develop their own data sets. These
are built on DLGs or other widely available products, which are then mosaicked, edited, corrected,
and updated.
One way to develop digital maps of more complex landscape features, such as soils, landforms,
geology, or vegetation types, is to digitize existing
maps. This is especially useful for small-scale relatively static features, such as geology, and for variables with source maps developed through tremendous expenditure of resources and input of
expertise, such as soils. A problem with digitizing
these types of maps is that they were probably developed using a traditional cartographic approach
and a "communicative paradigm" (DeMers, 1991),
through which the purpose of the map was to present a visual model of the spatial distribution of the
phenomenon at a particular scale (Goodchild,
1988). The data used to develop this model were
probably messy or noisy and have been smoothed
using cartographic methods. However, in GISbased analysis, we probably want to know the spatial distribution of the phenomenon with more precision and less smoothing, as well as with some
representation of the spatial distribution of the variance of the phenomenon or uncertainty in the map
(discussed later). Also, hard-copy source maps
were produced at a particular map scale that can be
generalized to a coarser scale in the GIS, but cannot be made finer without implying false precision.
Alternatively, over the last few decades, mapped
data have increasingly been developed for the first
time in digital form, especially using remotely
sensed data. These include vegetation and land
cover data at global and regional scales (Tucker et
aI., 1985; Townshend et aI., 1987; Loveland et aI.,
Geographic Infonnation Science and Ecological Assessment
TABLE 11.1. Geospatial data types and sources needed for developing vegetation data for ecosystem assessment.
Theme
Scale,
description
Source
Use
Processing
and analysis
Landsat Thematic
Mapper imagery
30 m
NASA, EROS
Data Center (via
Forest Service)
Identify
vegetation
formations and
land cover
Unsupervised
classification,
segmentation
Resource air
photos
Variable;
Forest Service
Conduct field
reconnaissance
and label
vegetation map
features
Aerial photo
interpretation
1 : 12,000 to
I : 30,000
Digital elevation
models (DEMs)
30-m, 7.5'
quadrangles
USGS (via
Forest Service)
Model potential
vegetation
series
Mosaic, smooth,
derive slope, aspect
Cartographic
Feature Files
(CFFs): land
ownership, forest
boundaries,
hydrology, roads
Variable; usually
1: 24,000
Forest Service
(from USGS
digital line
graphs and
various digitized
sources)
Isolate study
area, develop
draft hard-copy
maps for field
checking
Overlay,
georeference, edit
Existing digital
thematic maps
(forest plantations,
previous vegetation
maps)
Variable; usually
1: 24,000
Forest Service
Label certain
features in
vegetation map
(plantations)
Overlay, map editing
Example from a recently completed project to map existing vegetation on Forest Service lands in Southern California carried out
at San Diego State University (Franklin and Woodcock, 1997; Franklin et al., 2000).
tor form as digital line graphs (DLGs). These represent fundamental data for an EA involving any
type of cartographic modeling (discussed later) for
which information such as the distribution of a resource by land ownership or in proximity to roads
is needed. However, as with the raster DEMs discussed previously, these vector data are only as
good as the maps from which they were derived
(which may be outdated and contain errors) and the
quality of the digital conversion process (never free
from errors). For example, in our work in southern
California, we found that the stream network represented in the DLG contained some errors in the
location of streams when compared to georeferenced satellite imagery, but, more importantly,
lacked information on lower-order (intermittent,
seasonal) streams associated with important riparian ecosystems (Franklin et aI., 2000). Owing to
the importance of accurate digital maps of such basic features as land ownership and land use (wilderness, for example) for land management planning
and EA, many federal, state, and regional agencies
are beginning to develop their own data sets. These
are built on DLGs or other widely available products, which are then mosaicked, edited, corrected,
and updated.
One way to develop digital maps of more complex landscape features, such as soils, landforms,
geology, or vegetation types, is to digitize existing
maps. This is especially useful for small-scale relatively static features, such as geology, and for variables with source maps developed through tremendous expenditure of resources and input of
expertise, such as soils. A problem with digitizing
these types of maps is that they were probably developed using a traditional cartographic approach
and a "communicative paradigm" (DeMers, 1991),
through which the purpose of the map was to present a visual model of the spatial distribution of the
phenomenon at a particular scale (Goodchild,
1988). The data used to develop this model were
probably messy or noisy and have been smoothed
using cartographic methods. However, in GISbased analysis, we probably want to know the spatial distribution of the phenomenon with more precision and less smoothing, as well as with some
representation of the spatial distribution of the variance of the phenomenon or uncertainty in the map
(discussed later). Also, hard-copy source maps
were produced at a particular map scale that can be
generalized to a coarser scale in the GIS, but cannot be made finer without implying false precision.
Alternatively, over the last few decades, mapped
data have increasingly been developed for the first
time in digital form, especially using remotely
sensed data. These include vegetation and land
cover data at global and regional scales (Tucker et
aI., 1985; Townshend et aI., 1987; Loveland et aI.,