It’s been a while since I focused on the environment side of the whole “global change” thing that this blog is supposed to be covering . . . at least directly.  Pretty much everything we do in development is connected to the environment – indeed, of late I have been referring to climate change as development’s Kevin Bacon while at work: I can get you from climate change to a development challenge, or vice versa, in three steps or less.  But I have not been writing much on the subject directly.

However, thanks to Garry over at Resilience Science, I’ve just read a really interesting article in Science (and a nice counterpoint to the recent bin Laden ambulance chasing in that journal) by Steve Carpenter and a bunch of others on Early Warnings of Regime Shifts in ecosystems.  For years, I have been teaching my students about the challenges of global environmental change, and trying to impress upon them that the part of these changes I find most worrying are the parts that are hardest to predict – the thresholds when particular biophysical systems might make sudden, discontinuous transitions to new states.  What has worried me, and I think much of the global change community, the most is the fact that we are not sure where these thresholds are, nor are we sure what it looks like when we approach one.  Thus, there is a pervasive concern within the community that we won’t know we’ve crossed a threshold or done something irreversible.

Carpenter and his co-authors, however, tested the hypothesis that “catastrophic ecological regime shifts may be announced in advance by statistical early-warning signals such as slowing return rates from perturbation and rising variance” by artifically inducing a regime shift in an aquatic food web (Carpenter is a limnologist – he does lakes, as it were) while monitoring a nearby similar lake as a control.  Their finding: they could see statistical warnings of an impending regime shift for more than a year before it occurred, validating their chosen early warning indicators (chosen from previously constructed understandings of the food web in question, and a bit much to synthesize here).

That there might be early warning indicators, or that the variables chosen by Carpenter, et al served as useful early warning indicators for regime change in this particular system are not terribly surprising.  What is interesting, though, is that the authors were able to demonstrate in a real-world (experimental) context (as opposed to desk theorization) that the early warning signals of regime shift are in fact detectable and measurable.  Granted, this is for a small, bounded food web – but the demonstration is important in a much wider way.  If we can find early warning indicators for regime shift in a small food web, there is no reason why we cannot find indicators for other complex systems – we can find a lot more early warning indicators of the discontinuous changes we fear, and in enough time to possibly address those changes before they occur.

But one big caveat here: this study did not reveal the actual mechanisms of regime shift.  As the authors note:

The precise mechanism of the nonlinear transitions is not known for our experiment; it could be one of the processes proposed in the literature, or something else. These early warning signals are expected to occur for a wide class of nonlinear transitions (7). Even though the mechanism is not known, manipulation of an apex predator triggered a nonlinear food web transition that was signaled by early warning indicators more than a year before the food web transition was complete. Thus the early warning indicators appear to be useful even in cases where the form of the potential regime shift is not known.

It seems to me that there is a serious risk of conflating correlation and causation here – that the authors got a bit lucky in this experiment, but that in other systems without an adequate understanding of the mechanisms of change, false correlations could cause us to lose the signal of regime shift in the noise of inappropriate data points.  I’m not sure how, or if, they intend to address this . . . but I think they will have to, if we are to usefully apply this to our food-producing ecosystems in a manner that allows us to think about sustainable development and food security in a meaningful way . . .