94
General Data Collection and Sampling Design Considerations
of a sampling design. However, in practice, IREAs
have multiple objectives, which increases the difficulty of choosing the type of sampling design and
its properties. The following guidelines should be
followed wherever possible (Legendre and Legendre, 1998):
1. The sampling grain should be larger than a unit
object (e.g., an individual) and the same as or
smaller than the structure to be detected by the
design (e.g., a patch).
2. The sampling interval should be smaller than the
structure to be detected.
3. The sampling (characterization) extent should
be the same as the total area covered by the patterns and processes of interest.
The sampling grain and extent define the observation window in spatial pattern analysis. No structure can be detected that is smaller than the grain
or larger than the extent. The sampling (characterization) extent of an IREA may differ from the assessment area to effectively cover all patterns and
processes of interest. In the Interior Columbia
Ecosystem Management Project (ICBEMP), the
characterization area was larger than the initial assessment area defined by the objectives (Hann et
aI., 1997; Jensen et aI., 1997; Quigley and Arbelbide, 1997). As a consequence of the hierarchical
structure of ecosystems (Dale and O'Neill, 1999;
also see Chapter 2), more than one scale may be
relevant to the patterns and processes of interest.
For example, species diversity relationships change
with changes in scale (Stoms, 1991; Stoms and
Estes, 1993; He et al., 1994), and therefore diversity data should be collected in a manner that reflects this scale dependence. Data at multiple scales
are often required to characterize the vegetation of
an area (e.g., Franklin and Woodcock, 1997). In addressing the issue of scale, explicit consideration
should be given to the need for data aggregation
and its impact on the spatial properties of data sets
(e.g., error propagation, sources of uncertainty).
7.3 Use of Existing Data
Existing data are utilized in IREAs either alone or
integrated with de novo data for two reasons: (1)
IREA objectives require the analysis of data on past
conditions (e.g., for trend analysis), and (2) the cost
and time involved in gathering all needed data by
means of a de novo regional survey are likely to be
prohibitive. Therefore, the kinds of data that are already available must be determined, along with
their usefulness and cost effectiveness for a specific purpose, even if biases in such data exist
(Austin, 1991a; Bourgeron et aI., 1994a; Davis,
1995). Existing data may also be used to guide the
sampling design for new surveys as needed. The
use of existing data often involves compilation and
collation of disparate and interdisciplinary data. Problems must be solved concerning differences in sampling dates, sampling design, data storage, spatial representation, spatial resolution, and data quality (Davis
et al., 1991; Davis 1995). These differences may prevent the use of the data for some purposes. For example, some data may be suitable for hypothesis
testing, while others may be suitable only for descriptive analyses.
However, existing data that are well organized
in an integrated regional database (see Chapter 8)
may be a cost-effective approach for IREAs. Many
such collections of data have been compiled (e.g.,
the California Environmental Resources Evaluation System, CERES, discussed in Davis, 1995;
also see Chapter 32: Southern Appalachian Assessment, Southern Appalachian Man and Biosphere (SAMAB), 1996; also see Chapter 34:
ICBEMP, Gravenmier et aI., 1997) or are being collected under general provisions for wide access
(e.g., for digital data in the United States, U.S. Executive Office of the President, 1994; in the United
Kingdom, the National Geospatial Database, Nanson et aI., 1995).
7.3.1 Geographic Information Systems
Data Layers
Geographic information systems (GIS) are important components of IREAs (see Chapter 11). One
main advantage of GIS is the ability to use and interpret a variety of data sources (Treweek, 1999).
Completed IREAs, such as the ICBEMP (see Chapter 34) and the Sierra Nevada Ecosystem Project
(SNEP, 1996), are sources of many readily available regional and subregional data layers. The sheer
number and variety of the data layers produced during a regional assessment make them appealing
sources of information for analyses (e.g., more than
170 data layers and 20 associated databases for the
ICBEMP). However, any use of data layers from
mUltiple sources warrants caution because of varying scale, attributes, metadata completeness, and
the like (Davis et aI., 1991; Davis, 1995; Treweek,
1999; also see Chapter 11). During an IREA, problems may be encountered, such as incompatible
map legends and different definitions of the same
legend item, in combining multiple fine-scale maps
of an attribute into a single coarse-scale data layer.
General Data Collection and Sampling Design Considerations
of a sampling design. However, in practice, IREAs
have multiple objectives, which increases the difficulty of choosing the type of sampling design and
its properties. The following guidelines should be
followed wherever possible (Legendre and Legendre, 1998):
1. The sampling grain should be larger than a unit
object (e.g., an individual) and the same as or
smaller than the structure to be detected by the
design (e.g., a patch).
2. The sampling interval should be smaller than the
structure to be detected.
3. The sampling (characterization) extent should
be the same as the total area covered by the patterns and processes of interest.
The sampling grain and extent define the observation window in spatial pattern analysis. No structure can be detected that is smaller than the grain
or larger than the extent. The sampling (characterization) extent of an IREA may differ from the assessment area to effectively cover all patterns and
processes of interest. In the Interior Columbia
Ecosystem Management Project (ICBEMP), the
characterization area was larger than the initial assessment area defined by the objectives (Hann et
aI., 1997; Jensen et aI., 1997; Quigley and Arbelbide, 1997). As a consequence of the hierarchical
structure of ecosystems (Dale and O'Neill, 1999;
also see Chapter 2), more than one scale may be
relevant to the patterns and processes of interest.
For example, species diversity relationships change
with changes in scale (Stoms, 1991; Stoms and
Estes, 1993; He et al., 1994), and therefore diversity data should be collected in a manner that reflects this scale dependence. Data at multiple scales
are often required to characterize the vegetation of
an area (e.g., Franklin and Woodcock, 1997). In addressing the issue of scale, explicit consideration
should be given to the need for data aggregation
and its impact on the spatial properties of data sets
(e.g., error propagation, sources of uncertainty).
7.3 Use of Existing Data
Existing data are utilized in IREAs either alone or
integrated with de novo data for two reasons: (1)
IREA objectives require the analysis of data on past
conditions (e.g., for trend analysis), and (2) the cost
and time involved in gathering all needed data by
means of a de novo regional survey are likely to be
prohibitive. Therefore, the kinds of data that are already available must be determined, along with
their usefulness and cost effectiveness for a specific purpose, even if biases in such data exist
(Austin, 1991a; Bourgeron et aI., 1994a; Davis,
1995). Existing data may also be used to guide the
sampling design for new surveys as needed. The
use of existing data often involves compilation and
collation of disparate and interdisciplinary data. Problems must be solved concerning differences in sampling dates, sampling design, data storage, spatial representation, spatial resolution, and data quality (Davis
et al., 1991; Davis 1995). These differences may prevent the use of the data for some purposes. For example, some data may be suitable for hypothesis
testing, while others may be suitable only for descriptive analyses.
However, existing data that are well organized
in an integrated regional database (see Chapter 8)
may be a cost-effective approach for IREAs. Many
such collections of data have been compiled (e.g.,
the California Environmental Resources Evaluation System, CERES, discussed in Davis, 1995;
also see Chapter 32: Southern Appalachian Assessment, Southern Appalachian Man and Biosphere (SAMAB), 1996; also see Chapter 34:
ICBEMP, Gravenmier et aI., 1997) or are being collected under general provisions for wide access
(e.g., for digital data in the United States, U.S. Executive Office of the President, 1994; in the United
Kingdom, the National Geospatial Database, Nanson et aI., 1995).
7.3.1 Geographic Information Systems
Data Layers
Geographic information systems (GIS) are important components of IREAs (see Chapter 11). One
main advantage of GIS is the ability to use and interpret a variety of data sources (Treweek, 1999).
Completed IREAs, such as the ICBEMP (see Chapter 34) and the Sierra Nevada Ecosystem Project
(SNEP, 1996), are sources of many readily available regional and subregional data layers. The sheer
number and variety of the data layers produced during a regional assessment make them appealing
sources of information for analyses (e.g., more than
170 data layers and 20 associated databases for the
ICBEMP). However, any use of data layers from
mUltiple sources warrants caution because of varying scale, attributes, metadata completeness, and
the like (Davis et aI., 1991; Davis, 1995; Treweek,
1999; also see Chapter 11). During an IREA, problems may be encountered, such as incompatible
map legends and different definitions of the same
legend item, in combining multiple fine-scale maps
of an attribute into a single coarse-scale data layer.