7.2 Purpose of Survey, Data Attributes, and Properties of the Data Collection Process
93
complete enumeration of the units from which a
sample can be selected. However, such knowledge
of the sampling frame may not exist for many
ecosystem components. Furthermore, given the
multiple purposes of IREAs, sampling is seldom
directed at a single target, and an optimal sampling
strategy may be hard to define (Bunce et al.,
1996a). Finally, the scale of the information to be
collected may vary from object to object within a
sampling effort (e.g., sampling tree, grass, raptor,
and invertebrate species).
The broad multiscale, multitheme nature of
IREAs, in conjunction with the hierarchical structure of ecosystems, leads to four important questions regarding data collection for IREAs: (1) What
are the general data requirements of IREAs? (2)
What standard data collection procedures are appropriate? (3) What are the trade-offs among statistical theory, logistical practice, sampling efficiency, data for interpolation and extrapolation, and
specific applications? (4) Can existing data be used
alone or integrated with de novo data?
This chapter complements Chapters 5 and 6 by
describing pragmatic, applied aspects of the data
collection process for IREAs, including the use of
and deviations from classical sampling design, as
well as answers to the four questions listed above.
The chapter contains discussions of the purpose of
the data collection process and its impact on data
attributes and other properties of data sets, the use
of existing data, the need for representative data
and the trade-off between statistical theory and application, the choice of sample sites, the detection
of rarities, sample size and configuration, and interpolation and extrapolation from samples.
7.2 Purpose of Survey, Data
Attributes, and Properties of
the Data Collection Process
The first step in implementing a data collection
scheme for an IREA is to clearly formulate the specific uses of the data in the context of the assessment (Green, 1979; Dale and O'Neill, 1999; Treweek, 1999). The specific objectives of an IREA
define the data attributes and several other properties of the data sets. The data collection design can
meet the purpose of an IREA only if its objectives
are explicitly defined (Gauch, 1982; Austin, 1987;
1991b). The IREA objectives determine the applications for which the data will be used (Dale and
O'Neill, 1999). These fall into seven generic types,
each with its own, sometimes overlapping, data requirements: spatial variability characterization (see
Chapters 3 and 13), temporal variability characterization (see Chapter 19), assessment of past and
current ecological conditions (see Chapter 19),
species and community distribution patterns, biodiversity characterization, linkages between terrestrial and aquatic systems (see Chapter 22), and inventory and monitoring.
Decisions about the sources of data are central
to the development of a database for an IREA (see
Chapters 1, 5, and 8). Data collection can be very
costly when all aspects of the effort are considered:
development of relational and spatial databases,
quality control, standardization of data from different sources, and the like. Therefore, it is important to collect only data that are relevant to the issues at hand. Whether existing data, de novo data,
or both are collected, a formal sampling design
should guide the process.
The specific attributes of the data to be collected
vary according to the objectives. For example, the
data needed will differ depending on whether all
ecosystems or only the forested ecosystems of a region are characterized; whether, if only forests are
characterized, all successional stages or only
ecosystems of a certain age (e.g., old growth) are
considered; and whether the interest is in all taxa
(plants or animals) or a subset of taxa (e.g., trees
of a specific genus or granivorous birds):
IREAs involve spatial analysis of the patterns of
interest because these patterns are structured by
forces that have spatial components. The objectives
define the spatial (and temporal) scale of several
properties of a spatial sampling design (Wiens,
1989; Allen and Hoekstra, 1992; He et aI., 1994;
Legendre and Legendre, 1998; Dale and O'Neill,
1999):
1. Grain, the size of the elementary sampling unit.
The resolution (Schneider, 1994) of an assessment is equal to the grain size of the sampling
design.
2. Sampling interval, the average distance between
neighboring sample units (i.e., the lag in time
series analysis).
3. Sampling extent, the total area to be characterized by the assessment (see Chapter 1) or the
duration of a time series. The sampling extent
is termed the range by Schneider (1994).
The scales identified for the data collection
process must satisfy the objectives of an assessment (see Chapter 1) and therefore address scales
appropriate for the patterns and processes of interest (see Chapter 3). A single-objective assessment
allows precise scaling of the specific components
93
complete enumeration of the units from which a
sample can be selected. However, such knowledge
of the sampling frame may not exist for many
ecosystem components. Furthermore, given the
multiple purposes of IREAs, sampling is seldom
directed at a single target, and an optimal sampling
strategy may be hard to define (Bunce et al.,
1996a). Finally, the scale of the information to be
collected may vary from object to object within a
sampling effort (e.g., sampling tree, grass, raptor,
and invertebrate species).
The broad multiscale, multitheme nature of
IREAs, in conjunction with the hierarchical structure of ecosystems, leads to four important questions regarding data collection for IREAs: (1) What
are the general data requirements of IREAs? (2)
What standard data collection procedures are appropriate? (3) What are the trade-offs among statistical theory, logistical practice, sampling efficiency, data for interpolation and extrapolation, and
specific applications? (4) Can existing data be used
alone or integrated with de novo data?
This chapter complements Chapters 5 and 6 by
describing pragmatic, applied aspects of the data
collection process for IREAs, including the use of
and deviations from classical sampling design, as
well as answers to the four questions listed above.
The chapter contains discussions of the purpose of
the data collection process and its impact on data
attributes and other properties of data sets, the use
of existing data, the need for representative data
and the trade-off between statistical theory and application, the choice of sample sites, the detection
of rarities, sample size and configuration, and interpolation and extrapolation from samples.
7.2 Purpose of Survey, Data
Attributes, and Properties of
the Data Collection Process
The first step in implementing a data collection
scheme for an IREA is to clearly formulate the specific uses of the data in the context of the assessment (Green, 1979; Dale and O'Neill, 1999; Treweek, 1999). The specific objectives of an IREA
define the data attributes and several other properties of the data sets. The data collection design can
meet the purpose of an IREA only if its objectives
are explicitly defined (Gauch, 1982; Austin, 1987;
1991b). The IREA objectives determine the applications for which the data will be used (Dale and
O'Neill, 1999). These fall into seven generic types,
each with its own, sometimes overlapping, data requirements: spatial variability characterization (see
Chapters 3 and 13), temporal variability characterization (see Chapter 19), assessment of past and
current ecological conditions (see Chapter 19),
species and community distribution patterns, biodiversity characterization, linkages between terrestrial and aquatic systems (see Chapter 22), and inventory and monitoring.
Decisions about the sources of data are central
to the development of a database for an IREA (see
Chapters 1, 5, and 8). Data collection can be very
costly when all aspects of the effort are considered:
development of relational and spatial databases,
quality control, standardization of data from different sources, and the like. Therefore, it is important to collect only data that are relevant to the issues at hand. Whether existing data, de novo data,
or both are collected, a formal sampling design
should guide the process.
The specific attributes of the data to be collected
vary according to the objectives. For example, the
data needed will differ depending on whether all
ecosystems or only the forested ecosystems of a region are characterized; whether, if only forests are
characterized, all successional stages or only
ecosystems of a certain age (e.g., old growth) are
considered; and whether the interest is in all taxa
(plants or animals) or a subset of taxa (e.g., trees
of a specific genus or granivorous birds):
IREAs involve spatial analysis of the patterns of
interest because these patterns are structured by
forces that have spatial components. The objectives
define the spatial (and temporal) scale of several
properties of a spatial sampling design (Wiens,
1989; Allen and Hoekstra, 1992; He et aI., 1994;
Legendre and Legendre, 1998; Dale and O'Neill,
1999):
1. Grain, the size of the elementary sampling unit.
The resolution (Schneider, 1994) of an assessment is equal to the grain size of the sampling
design.
2. Sampling interval, the average distance between
neighboring sample units (i.e., the lag in time
series analysis).
3. Sampling extent, the total area to be characterized by the assessment (see Chapter 1) or the
duration of a time series. The sampling extent
is termed the range by Schneider (1994).
The scales identified for the data collection
process must satisfy the objectives of an assessment (see Chapter 1) and therefore address scales
appropriate for the patterns and processes of interest (see Chapter 3). A single-objective assessment
allows precise scaling of the specific components