eLearning

recise methods die hard. Even today, a quarter-
century after publication of our studies, bottle experi-
ments are very common.
There appear to be problems with bottle-scale
experiments with zooplankton as well. Marshall
and Mellinger (1980) showed that although large
bottle experimentswith zooplankton replicated pre-
cisely, communities in bottles were less diverse than
those in LakeMichigan and in ELA Lake 223.
There are cases where such very small scales
appear to give useful results. Rudd and colleagues
(1988) were able to assess the potential for sulfate
reduction and denitrification in lakes from small
sediment cores, demonstrating their validity by
comparisons with results in reference lakes and
those experimentally acidified with sulfuric and
nitric acid. They were then able to demonstrate the
widespread importance of these processes in inter-
nal alkalinity generation, enabling a general model
to be developed (Kelly and others 1987).
INTERMEDIATE SCALES:MESOCOSMS
Large mesocosms, constructed of clear or translu-
cent polyethylene, have been used for over 30 years
(Goldman 1962). They were assumed to add more
realism than bottles, simply because they were
larger and deeper, allowing phenomena such as
vertical migration of zooplankton, and in the case of
mesocosms bottomed in sediments, exchange of
chemicals between mud and water. Mesocosms can
be replicated, allowing rigorous statistical designs.
However, my experience has been that even meso-
cosms 10min diameter, containing several hundred
cubic meters of water, cannot be extrapolated with
confidence to whole lakes unless carefully designed
to duplicate key features of ecosystems. In brief,
simply making an experiment larger does not make
it ecosystem scale, for as I shall illustrate, key
biogeochemical processes, habitats, and community
components may still not be included.
Physical Shortcomings
One of our earliest applications of large mesocosms
was to use them to try to predict physical phenom-
ena such as mixing and gas exchange in ELA lakes.
We determined early on that we were unable to use
heat-flux methods used for mixing in larger lakes,
for vertical transport of chemical substances in the
thermocline of ELA lakes was less than the molecu-
lar diffusion of heat (Hesslein and Quay 1973; Quay
1977; Quay and others 1980). Instead, we added
small amounts of radioactive tracers to estimate
mixing. To determine the reliability of mesocosms,
we deployed them in Lakes 224 and 227, where