A causal framework for anticipating and managing ecological regime shifts

Detecting regime shifts in ecological systems is crucial for anticipating changes and guiding management actions or ecosystem restoration. When ecosystems are trapped in an undesirable state, assessing their resilience can provide guidance for deciding how and when to intervene on system to trigger a shift to a more desirable state. By understanding how strongly the system resists change, one can better anticipate the type, intensity and timing of such interventions.

To design effective interventions, it is necessary to distinguish causal effects from correlations and determine how acting on a given driver can change the system’s resilience. We show that adopting a causal approach provides tools to measure the resilience of a system — thus how far a bifurcation point is — and the effect of interventions, contributing to better ecosystem management.

We illustrate this approach using the example of freshwater eutrophication, where lakes can shift from a clear to a turbid state (and vice-versa). Using observational data combined with knowledge of causal interactions, we outline a protocol to measure system resilience and anticipate the effects of interventions — such as nutrient reduction or biomanipulation — tailored to the current regime. For example, causal effect estimation can help answering questions such as: given the current state of my system, what would be the effect of e.g. removing big fishes from the lake during one month? Should I reduce the resilience of the turbid state beforehand in order for that intervention to be sufficient, by e.g. reducing the nutrient input?

The method is designed as a general tool for experts and can be applied across multiple ecosystems exhibiting tipping dynamics. It provides a framework based on explicitly specifying the causal graph linking system variables, identifying which variables can be intervened upon, estimating resilience from observational data, and selecting interventions that achieve a predefined management goal while accounting for associated costs.