Introduction - Climate Change Impacts on Freshwater Ecosystems - page 10

hatched and their changing characteristics and genome traced through a period
of environmental change (Mergeay et al . 2004).
A parallel approach to palaeoecological studies is to use space-for-time
investigation, where existing climate gradients provide different systems for
examination. The gradient from Greenland to Greece in Europe provides a wide
range of systems in which processes and food webs can be compared to predict
how they might change as temperatures increase (Moss et al . 2004; Meerhof
et al . 2007). There are, as with every approach, problems with this otherwise
attractive endeavour. Not only does climate change along the gradient, but so do
relief, geology and the intensity of human activity. Good design of observational
schemes can correct for these by stratified random sampling, but one major
source of variation, accidents of history, cannot. Glaciation and the nuances
of biogeography impose differences that can only be judged. A formerly
glaciated lake in Finland, with an Ice-Age-depleted, still recolonizing biota, may
not respond to temperature increase in the same way as a long-established
Mediterranean lake that may have been affected but not obliterated by the ice of
the glacial period, 20,000 years ago, even if the Finnish lake eventually becomes
as warm as the Mediterranean one now.
The next stage of investigation is to attempt to reproduce alleged effects
through experimentation. Experiments can reveal mechanisms because the
drivers of change can be controlled, and experimental designs and adequate
replication allow the study of several simultaneous drivers. Experiments are thus
potentially more powerful than comparative observations. They also compel the
creation of mechanistic hypotheses that force the experimenter to think through
the processes that are going on. But the scale of the experiment is important in
ecology. Whole-system experiments (Carpenter et al . 2001) (clear-felled versus
undisturbed sub-catchments of a forested river system, lakes subdivided by
curtains and parallel-engineered river channels) are ideal but liable to
pseudoreplication because the experiments are so expensive, and the subjects so
individual, that generally only one system can be handled at a time. In contrast,
experimental laboratory microcosms (Petchey et al . 1999) can be replicated
extensively but lack reality. The fashion of using micro-organism communities to
mimic large-scale systems (Benton et al . 2007) is attractive but perhaps mostly to
theoreticians.
The compromise is to use subsystems of real communities: mesocosms in lakes,
artificial river channels or plots in wetlands, or mesocosm tanks big enough to
contain all or almost all of the structures and food-web levels of a system (McKee
et al . 2000, 2002, 2003; Liboriussen et al . 2005). Usually 'almost all' is apposite,
for the top predators of a fish community need much more space than is possible
in replicable mesocosms, and the complete complexity of a natural system, which,
in rivers, for example, might involve interactions with large land mammals
(Terborgh 1988; Ripple & Beschta 2004) and tonnages of dead timber, is beyond
contemplation.
Another compromise is to do the experiments on simulated systems or models
using computer technology. This is, of course, the approach taken by the IPCC in
modelling future climate change. Per se it is relatively inexpensive, but the models
are reflections of the data input to them. If there are unsuspected factors involved,