74
eral Geographic Data Committee's Content Standardsfor Digital Geospatial Metadata (1998). Adequate records of the process from data collection
to preservation are every investigator's responsibility, including field technicians (data entry), data
managers and statisticians, and principal investigators.
5.4.3 Analyzing and Using Data
Software packages have made it easier to perform
many statistical operations, while making it more
difficult to evaluate the assumptions and algorithms
of individual tests. The primary responsibility of
the investigator is to realize if the assumptions are
met and to understand how the algorithms used influence the interpretation of the results. For example, some spreadsheet software programs have a
mathematical function for the standard error of the
mean for a population, rather than for a sample.
This results in an underestimation of the true variability about the mean. Different kriging programs
may produce wildly different "contour maps" from
the same biological data depending on the algorithms that they use for interpolation (Legendre and
Fortin, 1989; Stohlgren et al., 1997b).
Data often are simplified for display purposes.
For example, vegetation classes may be lumped to
accommodate storage or printer capabilities. It is
important to maintain and link detailed records of
data reduction procedures to map versions and their
metadata. Maps and data made available on the
Web should carry attached metadata.
5.4.4 Data Acquisition for Predictive
Ecosystem Models for
Decision Making
The real potential of data acquisition for ecological assessments lies in developing and validating
predictive ecosystem models (Buckley et aI., 1993;
Stohlgren et al., 1997c). Here, various scenarios
can be tested nondestructively. Conceptual and predictive models are an important step in avoiding
the preventable degradation of biotic resources and
populations of special concern. The models can
also be powerful tools in planning, teaching, training, and public outreach. GIS-based ecosystem
models provide an ideal tool for synthesizing existing information and making it useful to resource
managers. For example, the spatially explicit
ecosystem model Savanna (Coughenour, 1993) has
Data Acquisition
been coupled to the ARCIINFO GIS system by the
RAPiD/Arc development environment (Buckley et
aI., 1993) to investigate links between biodiversity
and ecosystem function. Such models are in use in
Elk Island National Park, Alberta, Canada (Buckley et aI., 1993) and Yellowstone National Park
(Coughenour and Singer, 1991).
5.5 Understanding the Quality and
Limitations of the Data
The usefulness of data is dependent on their quality (Kanciruk et aI., 1986). It is important to develop predetermined levels of accuracy and precision for field measurements (Messer et al., 1991;
Palmer et al., 1991). Sources of error must be anticipated, evaluated, and minimized or eliminated.
Besides systematic error (bias in measuring devices) and sampling error (bias in the methodology
employed), long-term landscape-scale measurements may have considerable spatial variability and
temporal variability; the "signal-to-noise ratio"
may be very small.
5.5.1 Sampling Design Considerations
Despite the detailed data sets about particular sites,
Berkowitz et aI. (1989) stress that results from
long-term study plots may be difficult to extrapolate because of (1) fundamental differences between the study system and surrounding areas; (2)
unknowable boundary conditions or intrusion of
unique events; (3) unknown uncertainty or bias in
results due to poor replication; (4) the nature of undedying processes and phenomena (i.e., different
perturbations may act on the system in different
ways through time); and (5) inappropriate methods
or poor data. This demonstrates that an important
but underrated phase of the ecological assessment
process is peer review of the experimental design.
Hurlbert (1984) suggested that about 70% of ecological studies suffered design or statistical problems.
Many of these problems, particularly pseudoreplication, are best alleviated early in a project. Often, in
science, we give more attention to the review of
manuscripts for publication than we do to the review of protocols and sampling designs. Rigorous
plans for data analysis must be developed prior to
data collection, understanding that some features of
the sampling design may adapt and evolve in the
first few years of the research program. That is, de-
eral Geographic Data Committee's Content Standardsfor Digital Geospatial Metadata (1998). Adequate records of the process from data collection
to preservation are every investigator's responsibility, including field technicians (data entry), data
managers and statisticians, and principal investigators.
5.4.3 Analyzing and Using Data
Software packages have made it easier to perform
many statistical operations, while making it more
difficult to evaluate the assumptions and algorithms
of individual tests. The primary responsibility of
the investigator is to realize if the assumptions are
met and to understand how the algorithms used influence the interpretation of the results. For example, some spreadsheet software programs have a
mathematical function for the standard error of the
mean for a population, rather than for a sample.
This results in an underestimation of the true variability about the mean. Different kriging programs
may produce wildly different "contour maps" from
the same biological data depending on the algorithms that they use for interpolation (Legendre and
Fortin, 1989; Stohlgren et al., 1997b).
Data often are simplified for display purposes.
For example, vegetation classes may be lumped to
accommodate storage or printer capabilities. It is
important to maintain and link detailed records of
data reduction procedures to map versions and their
metadata. Maps and data made available on the
Web should carry attached metadata.
5.4.4 Data Acquisition for Predictive
Ecosystem Models for
Decision Making
The real potential of data acquisition for ecological assessments lies in developing and validating
predictive ecosystem models (Buckley et aI., 1993;
Stohlgren et al., 1997c). Here, various scenarios
can be tested nondestructively. Conceptual and predictive models are an important step in avoiding
the preventable degradation of biotic resources and
populations of special concern. The models can
also be powerful tools in planning, teaching, training, and public outreach. GIS-based ecosystem
models provide an ideal tool for synthesizing existing information and making it useful to resource
managers. For example, the spatially explicit
ecosystem model Savanna (Coughenour, 1993) has
Data Acquisition
been coupled to the ARCIINFO GIS system by the
RAPiD/Arc development environment (Buckley et
aI., 1993) to investigate links between biodiversity
and ecosystem function. Such models are in use in
Elk Island National Park, Alberta, Canada (Buckley et aI., 1993) and Yellowstone National Park
(Coughenour and Singer, 1991).
5.5 Understanding the Quality and
Limitations of the Data
The usefulness of data is dependent on their quality (Kanciruk et aI., 1986). It is important to develop predetermined levels of accuracy and precision for field measurements (Messer et al., 1991;
Palmer et al., 1991). Sources of error must be anticipated, evaluated, and minimized or eliminated.
Besides systematic error (bias in measuring devices) and sampling error (bias in the methodology
employed), long-term landscape-scale measurements may have considerable spatial variability and
temporal variability; the "signal-to-noise ratio"
may be very small.
5.5.1 Sampling Design Considerations
Despite the detailed data sets about particular sites,
Berkowitz et aI. (1989) stress that results from
long-term study plots may be difficult to extrapolate because of (1) fundamental differences between the study system and surrounding areas; (2)
unknowable boundary conditions or intrusion of
unique events; (3) unknown uncertainty or bias in
results due to poor replication; (4) the nature of undedying processes and phenomena (i.e., different
perturbations may act on the system in different
ways through time); and (5) inappropriate methods
or poor data. This demonstrates that an important
but underrated phase of the ecological assessment
process is peer review of the experimental design.
Hurlbert (1984) suggested that about 70% of ecological studies suffered design or statistical problems.
Many of these problems, particularly pseudoreplication, are best alleviated early in a project. Often, in
science, we give more attention to the review of
manuscripts for publication than we do to the review of protocols and sampling designs. Rigorous
plans for data analysis must be developed prior to
data collection, understanding that some features of
the sampling design may adapt and evolve in the
first few years of the research program. That is, de-