96
General Data Collection and Sampling Design Considerations
7.3.4 Plot Data
Plot or transect data are commonly collected in routine biological surveys and monitoring programs.
When incorporated into a standardized database,
they provide a very cost effective tool for IREAs.
A limitation in integrating different sources of existing survey data within a regional database is that
the same biotic or abiotic information is rarely collected in each survey. However, if a minimum list
of attributes common to all surveys is established,
the data can be used for a variety of purposes
(Austin, 1991a; Jensen et al., 1994). The lack of a
common sampling design and differences in sample size may limit the use of these data. For example, over 37,000 vegetation plots were acquired
for the ICBEMP and incorporated into ECADS, an
integrated ecological data management and analysis system (Jensen et al., 1998). These plots provided data for many aspects of the ICBEMP, including deriving vegetation structural stages
(O'Hara et al., 1996), evaluating regional-scale potential vegetation map accuracy (Reid et al., 1995),
determining potential vegetation in subsample areas (Hann et al., 1997), and constructing predictive
models of species and community distributions
(Humphries and Bourgeron, in Jensen et al., 1997).
However, not all 37,000 plots were suitable for addressing all objectives because of differing levels
of data completeness, plot characteristics, and so
on. Other examples of the use of regional plot databases for land management and conservation planning are provided by Austin et al. (1983, 1984,
1990,1996) and Margu1es et al. (1987) in Australia.
Such regional databases can also be used to determine which geographic areas and/or parts of environmental gradients have not been sampled. Systematic reexamination of existing data can prevent
duplication of survey effort and make maximum
use of resources. Disadvantages in the use of existing data (e.g., cost of establishing a regional database, limitation due to a minimum list of attributes
at the regional scale) may often be offset by potential cost reduction and increasing effectiveness
of new surveys and by providing an initial regional
overview of phenomena of interest.
To assess the effect of grazing on rangeland in
the Sierra Nevada for SNEP, existing Parker range
condition and trend transects were acquired that
had been repeatedly resurveyed over five decades
in ten Sierra Nevada National Forests; the information was supplemented by resurveying 24 of the
transects (Menke et al., 1996). Limitations in the
use of Parker transect data were acknowledged,
such as insensitivity of the method to changes in
plant composition in areas with low plant cover and
known species identification problems; but the
transects comprised the only long-term widespread
range information available for the Sierra Nevada
(Menke et al., 1996).
Regional gradient analyses and modeling of
plant species and community distributions have recently been conducted using regional plot databases
assembled from a variety of existing sources. For
example, Austin et al. (1990, 1996) analyzed individual Eucalyptus species and species richness responses to environmental variables using a 7200plot data set covering 40,000 km 2 in southeast New
South Wales, Australia. At a minimum, plots contained presence and absence of tree species. They
also contained information on key environmental
variables or were georeferenced accurately enough
to attribute plots with such information. Similarly,
Leathwick (1995, 1998) analyzed the response of
New Zealand tree species to environmental variables using over to,500 forest plots. Ohmann and
Spies (1998) analyzed regional gradient relations
of woody plant communities in Oregon based on a
subset of a to,OOO-plot data set acquired from
USDA Forest Service and Oregon State University
sources. Regional gradient analyses were conducted for vegetation and plant functional types in
the ICRB using data from 19,000 plots (out of
37,000 plots acquired for the ICBEMP), which contained the appropriate information for the analyses
(e.g., complete species cover and environmental
data, appropriate plot size; Bourgeron et al., unpublished).
Regional plot databases can be integrated with
GIS and remotely sensed data to derive detailed
spatial environmental data for large areas. He et al.
(1998) used Thematic Mapper (TM) data in conjunction with field inventory data (Forest Inventory
and Analysis, PIA) (Hahn and Hansen, 1985;
Hansen et al., 1992) in a GIS environment. They
overlaid the satellite classification (adapted from
Wolter et al., 1995) with an ecoregional classification (Host et al., 1996) in northwestern Wisconsin.
The resulting classes represented the distributions
of dominant tree species. PIA data were used to
generate age-class distributions for dominant
species and the distributions of associated species
for each age class of the dominants. Finally, the
age classes of the dominant and associated species
were assigned to each pixel of the dominant species
map. Results were used to assess forest patterns
across regional landscapes and as input into LANDIS, a forest simulation model (Mladenoff et al.,
1996; He and Mladenoff, 1999; also see Chapter
18), to examine forest landscape change over time.
General Data Collection and Sampling Design Considerations
7.3.4 Plot Data
Plot or transect data are commonly collected in routine biological surveys and monitoring programs.
When incorporated into a standardized database,
they provide a very cost effective tool for IREAs.
A limitation in integrating different sources of existing survey data within a regional database is that
the same biotic or abiotic information is rarely collected in each survey. However, if a minimum list
of attributes common to all surveys is established,
the data can be used for a variety of purposes
(Austin, 1991a; Jensen et al., 1994). The lack of a
common sampling design and differences in sample size may limit the use of these data. For example, over 37,000 vegetation plots were acquired
for the ICBEMP and incorporated into ECADS, an
integrated ecological data management and analysis system (Jensen et al., 1998). These plots provided data for many aspects of the ICBEMP, including deriving vegetation structural stages
(O'Hara et al., 1996), evaluating regional-scale potential vegetation map accuracy (Reid et al., 1995),
determining potential vegetation in subsample areas (Hann et al., 1997), and constructing predictive
models of species and community distributions
(Humphries and Bourgeron, in Jensen et al., 1997).
However, not all 37,000 plots were suitable for addressing all objectives because of differing levels
of data completeness, plot characteristics, and so
on. Other examples of the use of regional plot databases for land management and conservation planning are provided by Austin et al. (1983, 1984,
1990,1996) and Margu1es et al. (1987) in Australia.
Such regional databases can also be used to determine which geographic areas and/or parts of environmental gradients have not been sampled. Systematic reexamination of existing data can prevent
duplication of survey effort and make maximum
use of resources. Disadvantages in the use of existing data (e.g., cost of establishing a regional database, limitation due to a minimum list of attributes
at the regional scale) may often be offset by potential cost reduction and increasing effectiveness
of new surveys and by providing an initial regional
overview of phenomena of interest.
To assess the effect of grazing on rangeland in
the Sierra Nevada for SNEP, existing Parker range
condition and trend transects were acquired that
had been repeatedly resurveyed over five decades
in ten Sierra Nevada National Forests; the information was supplemented by resurveying 24 of the
transects (Menke et al., 1996). Limitations in the
use of Parker transect data were acknowledged,
such as insensitivity of the method to changes in
plant composition in areas with low plant cover and
known species identification problems; but the
transects comprised the only long-term widespread
range information available for the Sierra Nevada
(Menke et al., 1996).
Regional gradient analyses and modeling of
plant species and community distributions have recently been conducted using regional plot databases
assembled from a variety of existing sources. For
example, Austin et al. (1990, 1996) analyzed individual Eucalyptus species and species richness responses to environmental variables using a 7200plot data set covering 40,000 km 2 in southeast New
South Wales, Australia. At a minimum, plots contained presence and absence of tree species. They
also contained information on key environmental
variables or were georeferenced accurately enough
to attribute plots with such information. Similarly,
Leathwick (1995, 1998) analyzed the response of
New Zealand tree species to environmental variables using over to,500 forest plots. Ohmann and
Spies (1998) analyzed regional gradient relations
of woody plant communities in Oregon based on a
subset of a to,OOO-plot data set acquired from
USDA Forest Service and Oregon State University
sources. Regional gradient analyses were conducted for vegetation and plant functional types in
the ICRB using data from 19,000 plots (out of
37,000 plots acquired for the ICBEMP), which contained the appropriate information for the analyses
(e.g., complete species cover and environmental
data, appropriate plot size; Bourgeron et al., unpublished).
Regional plot databases can be integrated with
GIS and remotely sensed data to derive detailed
spatial environmental data for large areas. He et al.
(1998) used Thematic Mapper (TM) data in conjunction with field inventory data (Forest Inventory
and Analysis, PIA) (Hahn and Hansen, 1985;
Hansen et al., 1992) in a GIS environment. They
overlaid the satellite classification (adapted from
Wolter et al., 1995) with an ecoregional classification (Host et al., 1996) in northwestern Wisconsin.
The resulting classes represented the distributions
of dominant tree species. PIA data were used to
generate age-class distributions for dominant
species and the distributions of associated species
for each age class of the dominants. Finally, the
age classes of the dominant and associated species
were assigned to each pixel of the dominant species
map. Results were used to assess forest patterns
across regional landscapes and as input into LANDIS, a forest simulation model (Mladenoff et al.,
1996; He and Mladenoff, 1999; also see Chapter
18), to examine forest landscape change over time.