Causal attribution in ecological systems: scaling from populations to communities

Understanding causal relationships in ecological systems is essential for predicting how populations and communities respond to environmental change. Over the last decade, interest has grown around causal attribution using long-term ecological time series, with convergent cross mapping emerging as a powerful method for causal discovery. This method is well-suited for nonlinear, coupled ecological systems and tests for causality by detecting whether a driver’s influence leaves a measurable signal within the time series of an affected variable. Given the presence of a causal process, we can predict the time series of the driver (i.e. cause) from the time series of the effect because of this measurable signal. Convergent cross mapping therefore systematically searches for these signals of the driver in the time series of the effect and quantifies whether a causal relationship exists based on how accurately we can predict the time series of the driver. Although convergent cross mapping has been effectively applied to population dynamics of diverse taxa like fish, grasses, and microbes, its application at the community scale remains less explored. We investigated how community-level metrics, including species richness and community abundance, behave under convergent cross mapping compared to individual populations. Using five decades of data from the North American Breeding Bird Survey, which includes annual multi-site bird counts, we examined the causal influence of temperature on both population abundances and community metrics. Our analyses revealed contrasting patterns: at the population level, longer time series increase the likelihood of detecting causal relationships with temperature, consistent with expectations for convergent cross mapping’s performance in recovering driver signals over time. However, at the community level, the probability of detecting temperature-driven causal effects decreases with longer time series. This divergence suggests potential ecological or methodological complexities when scaling from populations to communities and indicates that aggregating species into community metrics may mask species-specific causal signals driven by temperature. These findings highlight challenges in scaling causal inference methods from populations to communities, emphasizing the need for careful interpretation of community-level causal dynamics in the face of global change.