We have a new paper out this week on the prevalence of alternate states in community dynamics. The study is led by lab postdocs Pierre Gaüzère and Lars Iversen. The main question was: how predictable is the composition of a community when the environment changes? A classic idea, common in nearly all forecasting applications, is that the community should track the environmental change – either rapid, equilibrium responses, or through slower, lagged responses. Nevertheless, the state of the environment should be sufficient to predict the state of the community. Is this true? In a recent theory paper (Blonder et al. (2017) in Ecology Letters), we argued that more complex dynamics are possible, including alternate community states representing strong hysteresis in the underlying dynamics. These alternate states challenge our ability to predict the responses of communities to environmental change, as multiple possible futures may be possible.

These alternate states can be detected on a community response diagram, which is a parametric plot of time series of community state (e.g. inferred temperature) and time series of environmental state (e.g. site temperature). Any scenarios where there are more than one community state for a single environmental state, have ‘state numbers’ above unity and therefore represent alternate states.

However, it remained uncertain whether such exotic scenarios occurred in reality. To find out, we leveraged data for both bird and plant community responses to climate change, over both Anthropocene and Late Quaternary timescales. Data came from a range of paleoecological sources (pollen sediment records) and modern sources (repeat bird surveys).

When we examined these data using community response diagrams, we found that alternate states were possible in both the bird and plant communities – two examples are shown below.

And when we examined the several hundred communities for which we had time-series data available, we found that these alternate states were actually reasonably common, and occurred non-randomly across the landscape.

The overall implications of the work are that alternate states are possible and likely in ecology, and that predicting community response to environmental change may not be simple. These conclusions were only possible to make in light of the long time series data available to us, and suggest that contemporary applications, with limited temporal coverage, may not be able to accurately predict community responses. I find these results scientifically intriguing, and also very discouraging for the challenges we face in global change ecology. We are still wrestling with these findings, and are planning follow-up work to better understand how these dynamics can occur.

You can read the study in Frontiers in Ecology and Evolution or download a PDF directly.