Stochasticity, succession, and environmental perturbations in a fluidic ecosystem

Abstract

Unraveling the drivers of community structure and succession in response to environmental change is a central goal in ecology. Although the mechanisms shaping community structure have been intensively examined, those controlling ecological succession remain elusive. To understand the relative importance of stochastic and deterministic processes in mediating microbial community succession, a unique framework composed of four different cases was developed for fluidic and nonfluidic ecosystems. The framework was then tested for one fluidic ecosystem: a groundwater system perturbed by adding emulsified vegetable oil (EVO) for uranium immobilization. Our results revealed that groundwater microbial community diverged substantially away from the initial community after EVO amendment and eventually converged to a new community state, which was closely clustered with its initial state. However, their composition and structure were significantly different from each other. Null model analysis indicated that both deterministic and stochastic processes played important roles in controlling the assembly and succession of the groundwater microbial community, but their relative importance was time dependent. Additionally, consistent with the proposed conceptual framework but contradictory to conventional wisdom, the community succession responding to EVO amendment was primarily controlled by stochastic rather than deterministic processes. During the middle phase of the succession, the roles of stochastic processes in controlling community composition increased substantially, ranging from 81.3% to 92.0%. Finally, there are limited successional studies available to support different cases in the conceptual framework, but further well-replicated explicit time-series experiments are needed to understand the relative importance of deterministic and stochastic processes in controlling community succession.

Keywords: GeoChip; community assembly; disturbances; metagenomics; remediation.

Fig. 1.

Basic conceptual frameworks for modeling the relationships of succession to stochasticity, ecosystem characteristics, and environmental perturbations. (A) Fluidic ecosystem with nutrient input. (B) Fluidic ecosystem with disturbance. (C) Nonfluidic ecosystem with nutrient input. (D) Nonfluidic ecosystem with disturbance. In all these cases, the degree of the compositional stochasticity before a perturbation is arbitrary, just for illustrative purposes in a relative scale. In the groundwater communities examined here, deterministic processes played a larger role than stochastic processes before the perturbation (e.g., EVO injection). Thus, in scenario A, we assume that the communities before perturbations are more under deterministic control. Also, the time for succession is at an ecological time scale and is more or less arbitrary. It could be days, months, years, and decades (63), depending on particular ecosystems and the nature and intensity of perturbations.

Fig. 2.

Detrended correspondence analysis (DCA) of all detected functional genes, showing successional trajectories of the groundwater microbial communities in the seven downgradient wells at day 0 (preinjection), 4, 17, 31, 80, 140, and 269 as well the upgradient control well. Colors of symbols represent different days for sampling.

Fig. 3.

Dynamic changes of stochasticity during the succession of the groundwater microbial communities. The stochasticity is defined as the complement of the selection strength, a proportion of the differences between the observed total similarity and the null expected similarity divided by the total similarity.

Fig. 4.

Ordination of community composition by nonmetric multidimensional scaling (NMDS) based on the modified Raup–Crick dissimilarity metric (15). The communities that are closer together are more deviant from the null expectation, whereas the communities that are farther apart are less deviant from the null expectations. The data suggested that the communities at days 4, 17, 31, 80, and 140 were less different from null model expectation, whereas the communities at day 0, 269, and the control well were far more different from null model expectation.