| dc.description.abstract | Understanding the relationships between organisms and their environment is crucial for the conservation
of species and ecosystems all over the world. In ecology, relationships between organisms and their
environment are typically obtained from correlation-based analyses of observational data. However,
understanding ecological processes and developing strategies to maintain them requires knowledge on
causation rather than correlation. Since causality between organisms and their environment is difficult
to study using observational data, ecologists may need advanced computational techniques to identify
causal relationships. In particular, the field of causal discovery offers a smorgasbord of algorithms to
establish causal relationships from observational data. Nonetheless, causal discovery algorithms have
not yet been widely adopted in ecology. In this review, we (1) introduce the concepts of causality
and causal discovery algorithms, (2) review the use of causal discovery algorithms in ecology, and (3)
present a perspective on the use of causal discovery algorithms and the interpretation of causality in
ecology, illustrated by the field of animal movement research.
So far, ecological studies have used causal discovery algorithms to establish causal relationships
between organisms and meteorological factors, and to identify causal interactions between species. The
most popular causal discovery algorithm in ecology is convergent cross mapping, which requires time
series data. Ecological studies often study causal links to a single response variable of interest, whereas
fewer studies consider causal links between more variables and create a causal network. Notably,
ecological interpretation of causal links can be hampered by changes in the strength and effect (positive
or negative) of the causal relationship over time or differences between observations.
To improve the interpretation of causal links in ecology, we identify three concepts that should
be considered in ecological causality research: First, we argue for a network-based approach in which
causal links between multiple variables are considered, because this could identify variables that modulate
causal relationships between other variables. Second, we stress that causes may differ between
time scales. Third, individual heterogeneity may hamper the interpretation of causal relationships.
Overall, broader use of causal discovery algorithms in ecology can increase our understanding of the
relationships between organisms and their environment, which could in turn help to conserve ecosystems. | |
