eLearning

for all inputs and outputs, these are typically disre-
garded in mesocosms.
In summary, after three attempts and some rather
complicated corrections, we were able to get meso-
cosm experiments to duplicate the responses of
whole lakes. If we had not had the perspective
provided by whole-lake responses to fertilization,
we would probably have accepted the results of the
first set of mesocosm experiments at face value,
concluding that low N:P loading ratios had no effect
on blue-green dominance. Any resulting manage-
ment decision would have been seriously flawed.
Even in answering rather simple questions,meso-
cosms required complicated correction for the mag-
nitudes of several processes in order to simulate
events in whole lakes properly. Hesslein and col-
leagues (1980) and Schindler and coworkers (1980)
compared the results of radiotracer additions to
whole lakes and mesocosms. Results were rather
dissimilar, but the reasons could not be explained.
In a subsequent mesocosm experiment, Santschi
and colleagues (1986) developed a measurement
and modeling program to separate the effects of
particle settling and resuspension, adsorption/
desorption to suspended particles, stagnant film
thickness at the sediment–water interface, and ben-
thic mixing rates to explain removal rates. When
the different magnitudes of these processes in lakes
were accounted for, their models reconciled whole-
lake and mesocosm results.
In summary, even for rather simple physical and
chemical questions, considerations of complex differ-
ences in several physical and chemical processes
were required in order to design mesocosm experi-
ments that could be extrapolated to whole lakes. If
we are to rely on mesocosm experiments to make
accurate predictions about changes to lakes, it is
clear that we must first use lake experiments to
understand, and possibly correct for, the shortcom-
ings of mesocosms. As I shall discuss, incorporation
of biology further complicates matters.
Biological Shortcomings
Problems with incorporating whole communities in
mesocosms. The problem of incorporating higher
trophic levels in mesocosm experiments has been
noted by many (Carpenter and Kitchell 1988;
Schindler 1988; O’Brien and others 1992; Pace
1998). In oligotrophic lakes such as those at ELA,
lake trout or pike are present at densities of from
0.01 to 2 fish per 1000 m3, so incorporating their
effects into mesocosm experiments is all but impos-
sible. Thus,mesocosms experiments involvingmore
than three trophic levels cannot be done. Even
when the effects of predators on planktivores were
experimentally simulated, mesocosms did not pre-
dictwell the long-termeffects of predatormanipula-
tions on lower trophic levels (Vanni and Findlay
1990; Findlay and others 1994).
There are other components of aquatic communi-
ties that are difficult to incorporate in mesocosms.
Recent studies have shown the importance of
pelagic–benthic coupling to nutrient cycling or
plankton predation in a wide variety of small lakes
(Schindler and others 1996a; F. Wilhelm and D. W.
Schindler unpublished). Only a small proportion of
studies on lakes are done at night, so the activities of
nocturnally active organisms are often underappre-
ciated. Organisms that are benthic by day and
pelagic at night are unlikely to be properly repre-
sented in pelagic mesocosms. Even if they were
included, it is difficult to envision how to simulate
their movements properly between littoral and
pelagic zones, or between epilimnion and hypo-
limnion.
An example is useful to illustrate this point.
Alpine lakes in the Cascade Valley of Banff National
Park, Alberta have been sampled since the mid-
1960s. Gammarus lacustris is known to be an impor-
tant predator in these normally fishless lakes. Re-
cords based on daytime sampling show it to be
largely benthic, with occasional large specimens
caught in pelagic samples taken in profundal depths
of the lakes. As a result, we did not incorporate
Gammarus in the large sets of mesocosm experi-
ments that we performed as pilot studies for whole-
lake predator restoration (Paul and Schindler 1994;
Paul and others 1995).
Nocturnal sampling was done for the first time in
1995 (a 14-mile hike through some of the best
grizzly habitat in the park, frequent snow, strong
wind, and subfreezing temperatures even in July
are effective deterrents to nighttime work). It was
discovered that in fishless lakes, high densities of
Gammarus occupy near-surface waters of the pelagic
zone at night. Sizes indicate that they originate
largely in the littoral zone (that is, their migration is
largely horizontal rather than vertical). They feed
voraciously on Hesperodiaptomus and other pelagic
crustaceans, but also transfer benthic nutrients to
the pelagic during the night by excretion. As a
result, phytoplankton biomass in Snowflake Lake is
much higher than in nearby lakes which have no
large Gammarus populations (F. Wilhelm and D. W.
Schindler unpublished data).
Temporal problems. In my experience, temporal
constraints are a severe handicap when extrapolat-
ing most mesocosm experiments (and, for that
matter, many whole-lake experiments that are car-
ried out only for a few years) to lake management.