Research

Unraveling the time-varying nature of responses to disturbances

Extreme weather events such as heatwaves and storms are increasing in frequency and magnitude under global change. Understanding the effects of such pulse disturbances requires rigorous theory that accommodates changes in the ecological context of a given community. Yet, the typical theoretical recipe to study responses of populations to disturbances assumes that the system is at an equilibrium with static species interactions. In my work, I have been extending this theory to non-equilibrium population fluctuations, such as cyclic and chaotic dynamics, which occur in many ecological communities. We have shown that, because of time-varying species interactions, several properties of how communities and their species respond to perturbations can also change over time. Specifically, we have shown that the sensitivity of species abundances to pulse disturbances depends on the timing of the disturbance (Medeiros et al 2023 Ecology Letters). Thus, any given species can be more or less sensitive depending on the ecological context of the community.

We have also proposed a theoretical framework to scale up from the sensitivity of individual species (Medeiros et al 2023 Ecology Letters) to whole-community sensitivity (Cenci et al 2020 J. R. Soc. Interface) by taking into account how species respond in a correlated way to pulse disturbances (Medeiros et al 2023 Ecology). By connecting metrics of responses to disturbance at the species and the community levels, this framework allowed us to bridge different levels of biological organization under non-equilibrium population dynamics. An important strength of these metrics of responses to perturbations is that they can be inferred directly from population time-series data using nonparametric statistical inference tools. This has allowed us to apply our approaches and obtain insights for several systems, from rocky intertidal to plankton communities.

Forecasting shifts in community composition and dynamics under disturbances

In addition to extreme weather events, human activities are causing persistent disruptions such as warming and pollution. Ecology has a long history of studying such press disturbances and their impacts, which include shifts in the composition or dynamical regime of communities. One important impact concerns sudden shifts to alternative dynamical regimes (e.g., alternative stable states). Although several time-series indicators, such as variance and autocorrelation, can warn about an upcoming shift, they provide no information on post-shift dynamics. We have developed a data-driven approach that combines machine learning tools with information on the press disturbance driver to characterize how populations will fluctuate after a regime shift (Medeiros et al 2024 bioRxiv). By applying this approach to an experimental microbial community and a natural planktonic community, we have shown that we can accurately reveal previously unseen dynamical regimes in these systems.

We have also developed ways to predict changes in community composition (i.e., which species will persist) under press disturbances. To do so, we have relied on the concept of structural stability, which captures the capacity of a community to maintain its composition under changes in the parameters of a population dynamics model (e.g., Lotka-Volterra model). By measuring the structural stability of all species subsets that can emerge from a species pool, we have shown that the subsets most likely to be assembled are those that can tolerate a larger range of press disturbances (Medeiros et al 2021 American Naturalist). This prediction has been confirmed using data on natural communities of herbivorous insects. We have also developed a theoretical framework that connects the response of a community to pulse disturbances and its response to press disturbances (Medeiros et al 2021 Journal of Animal Ecology). This framework allows us to obtain additional information about how a community should respond to perturbations by measuring a single metric from data.