Environmental Engineering Reference
In-Depth Information
Unfortunately, true experiments are infrequently conducted in ecological and
wetland studies because they are not feasible or would exceed available funding.
Frequently in wetland studies, the investigator has no control of the experimental
units or treatments (observational study). In these instances, true replication and
randomization are compromised in some way. The results from these studies are not
as conclusive as for experiments, but they do have merit and inference is possible as
long as the context of the study is fully revealed.
Assessment studies
usually have a single experiment unit of interest (i.e., a single
wetland), where treatments are randomized to test effectiveness in achieving a
conservation goal. For example, a manager may be interested in the most effective
technique to prevent amphibians from entering a contaminated wetland. The
hypothesized techniques (e.g., fence, moat, vegetation removal) form the treatments
and are randomly applied to the single wetland with a response variable of number of
amphibians crossing the treatment into the wetland. Inference is only possible to the
single experimental unit being subjected to the treatments.
Observational
or
descriptive studies
are very common in wetland science.
Essentially, attributes of variables of interest are measured in multiple experimental
units over space and time to describe what is observed. Commonly used to develop
monitoring strategies or support conclusions from retrospective studies, observa-
tion studies are designed to describe the systems of interest and, through the use of
retroduction, suggest causal relationships for the measured variables. In an obser-
vational study, the system is not manipulated so that variables of interest are not
isolated such as in true experiments. The primary drawback to observation studies
is that any one of many potential causes could have resulted in the measured
observation. Therefore, it is imperative to have considerable understanding of the
wetland type being studied to develop a plausible explanation for the observed
patterns. Evidence supporting strong conclusions from observational studies is
usually lacking, which prompted Romesburg (
1981
) to urge researchers to use
results from observational studies as hypotheses subjected to more rigorous
experiments to test the tentative conclusions.
Treatments in an experiment can be categorized into three types.
Manipulative
treatments
refer to those studies where all experimental units have the same
probability of being randomly assigned to any of the treatments in the study. An
example of manipulative treatments would be the random assignment of soil
moisture manipulations (e.g., flood, dry, moist) among all potential wetlands
under study to measure the production of certain plant communities. Compared
with other treatment types, studies with manipulative treatments have the greatest
scientific rigor or reliability of results.
Organismic treatments
are defined when
experimental units are a treatment, usually categorical, by definition. That is, it is
impossible to randomly assign treatments and it is assumed that experimental units
represented by these treatments are representative of a random sample from the
target population. Examples of such categories include sex and age of animals,
plant species or type, soil series, and wetland type. The third type of treatment,
which is frequently important for studies with a temporal factor, is the
repeated
measure
of experimental units throughout the study, where time is considered a
