It’s a truth universally recognised that any good field ecologist is in search of a good control. Since most studies of human impact on ecosystems are ad hoc the history of control sites is often unknown and can lead to problems with interpretation, particularly in fields like invasion biology.
However it gets worse. Treatments in ecology are often not even really treatments.
In the lab most biologists can strictly control what each treatment constitutes (damn them) but in the field this is often not possible. As such there can be important differences in the treatment we are trying to characterise. This problem was already noted more than two decades ago by Michael Huston who stated that: “When an experimental manipulation has multiple components, but only one of them is identified as the experimental treatment, erroneous conclusions about cause and effect relationships are likely because the actual cause of any observed response may be ignored in the interpretation of the experimental results.”
Some work I have been doing during my thesis made me realise that this problem is actually still extremely common in certain ecological disciplines. I’ve noticed that work on logged forests and invasive species seem to be particularly prone to this (but I am certain that there are many other fields in which this regularly happens).
In studies of logging in tropical forest it is common to make a statistical comparison between logged and unlogged forests. However, not all logged plots are the same and there tends to be a very large variation in logging damage even in individual concessions. Despite this logging studies often analyse changes in forest characteristics using t-tests and lump together all logged forests as if they were the same. Doing this ignores lots of interesting and important variation that could be vital to management of tropical forests.
Similarly in studies of invasive plant species comparisons are commonly made between uninvaded and invaded ecosystems. However, again there is usually lots of variation regarding the degree of invasion within individual sites. There has been lots written on different hypotheses regarding the impact of invasive species, one of which is that the impact is directly linked to the abundance/biomass of species. Simply by making comparisons between invaded and uninvaded ecosystems will not answer this question, we need to look at how impact varies with abundance. Addressing this question may give us some evidence on the degree to which invasive species are drivers or passengers of degradation – one of the most contentious questions in invasion biology.
The stupid thing is that these problems could be fixed relatively easily if we viewed these treatments as gradients of change. This allows a more nuanced interpretation of the potential impacts of humans on ecosystems changing the question from “Does impact x cause a change in y?” to “How does variable y change as impact x changes?” (Which Brian McGill wrote on over at Dynamic Ecology a while back). This question is infinitely more interesting and more informative if we want ecology as a field to get better at prediction. Very few human impacts can truly be seen as a treatment so we should stop treating them as if they were.