Study Design and Logistics - Wetland Techniques - page 23

radiating from the core location of disturbance. The resulting data are used to model
the decline of impact effect with linear distance.
Hannaford and Resh ( 1999 ) used a BACI study design to estimate the impact of
all-terrain vehicles (ATV) on vegetation in a San Francisco Bay wetland. They found
that ATV use caused immediate impact to vegetation but limited use allows for
recovery within a year without continuing traffic. Zimmer et al. ( 2001 )usedaBACI
approach to assess the ecological response to colonization and extinction of minnows
in a prairie pothole in Minnesota. The impacted wetland was paired with a fishless
site and comparisons were made when both were fishless, following introduction of
fish into impact wetland, and after eradication of fish in impacted wetland. They
found that introduction of fish into a prairie pothole resulted in increased turbidity,
total phosphorus and chlorophyll a in water, and decreased abundance of aquatic
insects. Removal of fish reversed these effects. Suren et al. ( 2011 )followedaBACI
protocol to evaluate the effect of hydrologic restoration of drains within a wetland in
New Zealand. Results indicated that restoration of drains was beneficial as inverte-
brate communities were similar to natural wetlands and cover of exotic pasture
grasses declined. In addition, connectivity was improved for recolonization of native
wetland plant and aquatic invertebrate communities.
1.9 Sampling
A consistent and common criticism of scientific studies is the scale to which study
results are applied beyond the target population (i.e., inference). It is rare to
measure every member of a target population (i.e., a census), which is why experi-
mental design and statistical analyses are crucial for study design. Therefore, a
subset of potential experimental units from the target population is usually selected
to measure the variables of interest, which is termed sampling . The selected
sampling design, as detailed later, is a contributing limiting factor of the extent of
inference from a study. In order for results to be extended to the target population,
the sampled experimental units must be representative of the target population (i.e.,
random selection and replication). Ultimately, statistical analyses of data collected
from an appropriate study design enable scientists to make inferences about a target
population from its sample.
Representative sampling of a target population allows for the description of
spatial and temporal patterns in nature and, through testing competing hypotheses,
linking an ecological process to the observed pattern. Ultimately, proper study
design should elucidate the linkages between described patterns and the ecological
processes that created the pattern. Unfortunately, few studies go beyond description
of a pattern with a conclusion based on retroductive speculation on the processes
that created the pattern. However, all investigators must realize that wetland data
are created by two classes of processes. The first is the ecological process that
generated the true pattern. The second is the process inherent in the sampling effort
that resulted in the data of interest. The assumption is that the sampling process