eLearning

months. Phytoplankton and zooplankton recovered
completely in the first year after fertilization ceased
(Shearer and others 1987).With respect to phospho-
rus–plankton systems, we believe that different
years are essentially uncorrelated, as borne out by
comparing the conclusions of using single years
versus long-term average responses for the several
lakes. Such fast-responding components are also
amenable to treatment by time-series analysis, as
long as periods of observation greatly exceed the
response times of the variables in question. Despite
the aforementioned long delays in responses of
chemical and biological components, it is difficult to
obtain funding for more than a few years, even for
large-scale experiments. Many of the arguments to
support long-term studies [for example, see Likens
(1992) and Edmondson (1991)] apply to experimen-
tal as well as to nonexperimental ecosystems.
One advantage of ELA experiments is that perma-
nent staff were hired to carry out the experiments.
Equipment was used for several experiments, and
accommodationwas available at the field camp. As a
result, individual experiments were not very costly
to begin. Also, experiments could be maintained in
the long term by reducing sampling scales. How-
ever, whole-ecosystem experiments can be more
costly where overhead, personnel, and transporta-
tion and equipment must be purchased. Carpenter
(personal communication) reports that lakemanipu-
lations in Wisconsin have required $350,000–
$600,000 per year. The European CLIMEX project
(climate change experiment), with its computer-
controlled gas and temperature regimes was even
more costly (R. F.Wright personal communication).
Whole-lake experimentswith nutrients and strong
acids have been carried out in several lakes and at
different sites. The results and conclusions have
differed only in minor details at lower trophic levels
(Carpenter and others 1991; Schindler and others
1978, 1991), in line with the expectation described
above. Also, the responses agreed well with those
observed in nonexperimental, eutrophied and acidi-
fied lakes (Schindler and others 1978, 1991). Simi-
larly, NITREX (nitrogen saturation experiments) at
several sites across Europe fit one general model for
nitrate release (Emmett and others 1998). These
observations suggest that perhaps extensive replica-
tion at one site is not absolutely necessary. However,
while lake-trout lakes responded similarly regard-
less of location, lakes with different fish assemblages
exhibited very different responses, regardless of
whether the lakes were in the same region (Schin-
dler and others 1991). Presumably, this was due to
the different sensitivity to acidification of different
fish species. It suggests that if several whole-
ecosystem experiments are possible, it may be more
informative to do them in ecosystems with different
species assemblages than to replicate them in nearly
identical systems.
There are also institutional constraints to whole-
ecosystem experiments in most areas. Few lakes or
other ecosystems in the USA or Europe are free of
conflicting interests, and it is difficult to have them
set aside as the exclusive preserve of researchers.
For example, permission for the Little Rock Lake
experiment required a special act of the Wisconsin
Legislature (T. M. Frost personal communication).
Even when permission for an experiment is ob-
tained, it is often impossible to control other uses of
the ecosystem that can confound the interpretation
of experiments. Many important perturbations are
‘‘off limits’’ even in dedicated whole-ecosystem re-
search facilities. For example, whole-lake experi-
ments with dioxins and nonnative species cannot be
done at ELA and are unlikely to be permitted
anywhere. Many case histories are available, how-
ever, for such ‘‘experiments’’ done accidentally or to
‘‘enhance’’ ecosystems. Carpenter (1998) discusses
many other conflicts that are faced by those who
would experiment at large scales. The bureaucracy
of dealing with such problems can be a significant
deterrent to young ecologists, anxious to demon-
strate their research prowess by publishing research
studies.
Experiments in previously perturbed ecosystems
may not reveal the entire range of responses that are
possible in unperturbed ones. In many cases, previ-
ous perturbations such as exploitation or introduc-
tion of nonnative species have already ‘‘peeled
away’’ natural biodiversity that may, at least in
theory, provide valuable protection for ecosystems,
in terms of ‘‘functional redundancy’’ (Schindler
1988, 1990; Frost and others 1995). The presence or
absence of natural species assemblages can affect the
results obtained in experiments at any scale. I have
written more extensively elsewhere on the pitfalls
of not knowing the true ‘‘baseline state’’ of ecosys-
tems (Schindler 1995).
CORRECTION FOR SPATIAL SCALES:
AN UNDERAPPRECIATED PROBLEM
At least some of the problems in scaling of meso-
cosm results to whole ecosystems also apply when
extrapolating from small lakes to large ones. Simple
extrapolation is valid in some cases. For example,
Schindler and colleagues (1978) showed that the
same sort of simple models dependent on phospho-
rus loading and water renewal worked well for both
large lakes and small. However, results from a study