98
General Data Collection and Sampling Design Considerations
along environmental gradients. The gradsect sampling design (Gillison and Brewer, 1985; Austin
and Heyligers, 1989) is intended to provide a costeffective description of the full range of biotic variability in a region by sampling along the full range
of environmental variability present. Transects that
contain the strongest environmental gradients in a
region are selected in order to optimize the amount
of information gained in proportion to the time and
effort spent during a survey (Austin and Heyligers,
1989). In addition, sampling sites are deliberately
located to minimize travel time. The method has
been shown statistically to capture more information than standard designs (Gillison and Brewer,
1985).
Mackey (1993) conducted a vegetation survey in
the wet tropics of Queensland, Australia, to examine prediction of rain forest structural characteristics from climate and soil parent material. A sampling design was required that would identify a
minimum number of sites that provided both representation and replication. Stratification by climate
zone, parent material, and topographic position, with
replication of most variable combinations, resulted
in a total of 61 plots for a preliminary analysis of
rain forest structure and physiognomic attributes.
A fine-scale gradsect vegetation survey was conducted in semiarid woodlands of eastern Australia,
in which gradsects were positioned along topographic gradients in three landscape types to measure vegetation, soil surface, and terrain attributes
in contiguous I-m 2 quadrats (Ludwig and Tongway, 1995). Gradsects ranging from 300 to 500 m
in length were subjected to boundary analysis to
characterize the spatial organization of each landscape at a range of scales.
Wessels et al. (1998) evaluated the effectiveness
of the gradsect method for conducting animal surveys compared to a comprehensive habitat-specific
survey in a 350-km2 nature reserve in South Africa.
Three 500-m-wide gradsects were positioned to incorporate the maximum variation in physiographic
characteristics. The gradsect method performed
well in detecting bird and dung beetle species compared to the more expensive and time-consuming
habitat-specific method. The authors concluded
that the gradsect method could be employed with
confidence for faunal surveys in areas where little
information is available about the vegetation. However, the gradsect method missed several species
associated with small, isolated patches of vegetation habitat that were not identified by the environmental variables used to construct the gradsects.
Therefore, vegetation information, where available,
can refine and improve a gradsect sample design
for animal surveys (Wessels et al., 1998).
7.4.3 Multistage Stratified
Semirandom Sampling
Steele (see Chapter 6) discussed the usefulness of
multistage stratified random sampling design in
which a second level of stratification is imposed on
a first level. This multistage approach has been
widely used in biological assessments (e.g., Debinski and Brussard, 1992; Debinski and Humphrey,
1997). Helman (1983) and Austin and Heyligers
(1989, 1991) have similarly expanded the gradsect
methodology to include different levels of environmental stratification within each gradsect, thereby
creating a multistage stratified semirandom sampling
design (MSSRS). Their modified gradsect procedure
utilizes a two-stage sampling design: (1) gradsects
are selected and (2) adequate environmental stratification and replication are performed within gradsects.
The two-stage gradsect sampling design has been
used to describe the rain forests of southern New
South Wales, Australia (Helman, 1983) and for a
mixture of eucalypt and rain forests in northern
New South Wales (Austin and Heyligers, 1989,
1991). Similar MSSRS methodology was used to
assess the conservation value of a 1300-km2 area,
the Gray Ranch, in southern New Mexico. The specific purpose of the survey was to characterize vegetation patterns and their associated floristic variability in relation to the range of environmental
variability within the ranch (Engelking et aI., 1994;
Bourgeron et aI., 1995), constrained by a two-week
time limit for fieldwork by four surveyors. Stratifying variables (geology, elevation, and soil type)
were chosen based on the results of a previous vegetation analysis. Taking access roads into consideration, a total of four gradsects was positioned to
contain 49 out of the 55 biophysical environments
(combinations of the stratifying variables) found at
the ranch. Sampling of the biophysical environments and of the gradsects was proportional to their
representation within the ranch. Geographic replication was provided by environmental overlap in
the gradsects, as well as by dividing each of the
two main gradsects into three segments. Actual
sample location was chosen randomly for a particular physical environment within each segment.
Another variant of the MSSRS-gradsect design
was utilized in designing a regional survey of
forest-dwelling microchiropteran bats in New
South Wales, Australia (Mills et aI., 1996). Temperature and precipitation attributes of forested areas were selected for environmental stratification
based on an understanding of the autecology of the
region's bat species. First, three areas were chosen
to represent climate strata. Second, four trapping
General Data Collection and Sampling Design Considerations
along environmental gradients. The gradsect sampling design (Gillison and Brewer, 1985; Austin
and Heyligers, 1989) is intended to provide a costeffective description of the full range of biotic variability in a region by sampling along the full range
of environmental variability present. Transects that
contain the strongest environmental gradients in a
region are selected in order to optimize the amount
of information gained in proportion to the time and
effort spent during a survey (Austin and Heyligers,
1989). In addition, sampling sites are deliberately
located to minimize travel time. The method has
been shown statistically to capture more information than standard designs (Gillison and Brewer,
1985).
Mackey (1993) conducted a vegetation survey in
the wet tropics of Queensland, Australia, to examine prediction of rain forest structural characteristics from climate and soil parent material. A sampling design was required that would identify a
minimum number of sites that provided both representation and replication. Stratification by climate
zone, parent material, and topographic position, with
replication of most variable combinations, resulted
in a total of 61 plots for a preliminary analysis of
rain forest structure and physiognomic attributes.
A fine-scale gradsect vegetation survey was conducted in semiarid woodlands of eastern Australia,
in which gradsects were positioned along topographic gradients in three landscape types to measure vegetation, soil surface, and terrain attributes
in contiguous I-m 2 quadrats (Ludwig and Tongway, 1995). Gradsects ranging from 300 to 500 m
in length were subjected to boundary analysis to
characterize the spatial organization of each landscape at a range of scales.
Wessels et al. (1998) evaluated the effectiveness
of the gradsect method for conducting animal surveys compared to a comprehensive habitat-specific
survey in a 350-km2 nature reserve in South Africa.
Three 500-m-wide gradsects were positioned to incorporate the maximum variation in physiographic
characteristics. The gradsect method performed
well in detecting bird and dung beetle species compared to the more expensive and time-consuming
habitat-specific method. The authors concluded
that the gradsect method could be employed with
confidence for faunal surveys in areas where little
information is available about the vegetation. However, the gradsect method missed several species
associated with small, isolated patches of vegetation habitat that were not identified by the environmental variables used to construct the gradsects.
Therefore, vegetation information, where available,
can refine and improve a gradsect sample design
for animal surveys (Wessels et al., 1998).
7.4.3 Multistage Stratified
Semirandom Sampling
Steele (see Chapter 6) discussed the usefulness of
multistage stratified random sampling design in
which a second level of stratification is imposed on
a first level. This multistage approach has been
widely used in biological assessments (e.g., Debinski and Brussard, 1992; Debinski and Humphrey,
1997). Helman (1983) and Austin and Heyligers
(1989, 1991) have similarly expanded the gradsect
methodology to include different levels of environmental stratification within each gradsect, thereby
creating a multistage stratified semirandom sampling
design (MSSRS). Their modified gradsect procedure
utilizes a two-stage sampling design: (1) gradsects
are selected and (2) adequate environmental stratification and replication are performed within gradsects.
The two-stage gradsect sampling design has been
used to describe the rain forests of southern New
South Wales, Australia (Helman, 1983) and for a
mixture of eucalypt and rain forests in northern
New South Wales (Austin and Heyligers, 1989,
1991). Similar MSSRS methodology was used to
assess the conservation value of a 1300-km2 area,
the Gray Ranch, in southern New Mexico. The specific purpose of the survey was to characterize vegetation patterns and their associated floristic variability in relation to the range of environmental
variability within the ranch (Engelking et aI., 1994;
Bourgeron et aI., 1995), constrained by a two-week
time limit for fieldwork by four surveyors. Stratifying variables (geology, elevation, and soil type)
were chosen based on the results of a previous vegetation analysis. Taking access roads into consideration, a total of four gradsects was positioned to
contain 49 out of the 55 biophysical environments
(combinations of the stratifying variables) found at
the ranch. Sampling of the biophysical environments and of the gradsects was proportional to their
representation within the ranch. Geographic replication was provided by environmental overlap in
the gradsects, as well as by dividing each of the
two main gradsects into three segments. Actual
sample location was chosen randomly for a particular physical environment within each segment.
Another variant of the MSSRS-gradsect design
was utilized in designing a regional survey of
forest-dwelling microchiropteran bats in New
South Wales, Australia (Mills et aI., 1996). Temperature and precipitation attributes of forested areas were selected for environmental stratification
based on an understanding of the autecology of the
region's bat species. First, three areas were chosen
to represent climate strata. Second, four trapping