Post provided by James T. Thorson

Ecologists have a social responsibility to document, interpret, and forecast how human activities are impacting our shared world.  There’s an ongoing movement to open ecology to new voices: for example, the Biden-Harris Administration has directed US agencies to incorporate Indigenous Knowledge through ethical and mutually beneficial relationships with tribal nations.  This direction is clearly important, and requires developing new ways to approach ecological learning.

Given this push, many ecologists are seeking to collaborate with local, traditional, and Indigenous knowledge-holders.  In some cases, this can result in new “data” (i.e., observations that are incorporated into quantitative models, or used to make regulatory decisions), and clearly this effort is positive when all communities feel that their input is treated respectfully.  However, treating local, traditional, and Indigenous knowledge as “data” results in theoretical and practical difficulties regarding how to weight these data relative to other observations.  This laudable goal could then result in generations of knowledge being overwhelmed by the larger “sample sizes” that are available to modern scientists.

In our recent study, we develop an alternative method (and software package dsem) that could be used to incorporate local, traditional, and Indigenous knowledge into ecological learning.  Specifically, we build upon work from famous ecologist Dr. Dick Levins (1930-2016), who foreshadowed the current push to make ecological modelling more inclusive.  Dr. Levins outlined how ecological analysis can “deal with complex systems as wholes” by specifying variables (boxes) and dependencies (arrows), and predicting system dynamics from these simplified “box-and-arrow” models.  We extend this work by showing how box-and-arrow models representing system interactions (“causal models”) can be fitted to a sequence of data (“time-series”) using efficient statistical methods.  Importantly, these box-and-arrow models might include rapid interactions that are essentially “simultaneous”, or slower (“lagged”) interactions where the impact of one variable on another only shows up after some time.  Ecosystems include both simultaneous and lagged interactions, e.g., where a predator might have a simultaneous and negative impact on prey (due to consumption), and prey might have a positive and lagged impact on predators (due to more energy to raise their young).  Estimating simultaneous and lagged interactions using time-series data then unifies a wide range of common ecological models.

Importantly, these new methods are fast enough that we can explore the impact of different assumptions about system dynamics in real-time, e.g., while holding workshops and in discussion with local, traditional, and indigenous knowledge-holders.  This new process for building ecological models then provides an alternative to treating knowledge as “data”; instead, generations of insight can be shared to inform the assumed linkages among system variables (i.e., arrows connecting boxes), building a shared understanding about how different ecosystem components are interconnected.

Critically, ecologists cannot conduct experiments on entire ecosystems.  Instead, we are restricted to smaller experiments on individual processes, which may or may not scale up when applied to whole-of-ecosystem dynamics.  However, our scientific ignorance at ecosystem scales provides an open door to local, traditional, and Indigenous knowledge-holders.  In these cases, scientists and knowledge-holders can be equal partners in hypothesizing different linkages, confronting them with data, and using them to explore the likely consequences of different policies.

Our recent paper compiled examples including trophic cascades for kelp forests, economic linkages during the Great Depression, overwinter survival for fish under climate change, and predator prey interactions for wolves and moose.  Since writing the paper, we have also collaborated to represent the impact of declining sea-ice in the Bering Sea ocean ecosystem (Fig. 1).  Going forward, we hope to collaborate more broadly, working to incorporate diverse viewpoints to better understand complex systems.

You can read the full article on Methods in Ecology and Evolution here