In a rapidly changing, non-stationary world it is imperative for natural resource managers to understand how biotic communities may respond to a range of environmental disturbances. Predicting how these disturbances may affect the overall stability of biotic communities is of particular interest and is shifting the historical emphasis on species-level dynamics to a new focus on species’ interconnections within large, complex communities. Network models are quickly becoming the ecologist’s tool-of-choice for studying multi-species interconnections within biotic communities. Network graphs, which represent species as ‘nodes’ and interconnections as ‘edges,’ are powerful yet intuitive tools for documenting and visualizing species’ connections at the system level. In addition, graph theory provides a mathematical framework for quantifying network structure and testing hypotheses on network assembly and stability; however, two key knowledge gaps currently limit the utility of the network approach. First, ecologists have so far focused on discrete mutualistic plant-pollinator or host-parasite interactions, or on antagonistic predator-prey interactions, but associations that are less easily classified (e.g., commensalism) or that are diffuse (i.e., realized through an intermediary) also can have strong influences on community structure and methods to incorporate them are now needed. Second, ecological network theory can predict how an acute extinction event (i.e., node removal) will influence the overall network, but it does not yet predict how ecological networks will respond to chronic environmental disturbances (i.e., those that do not cause acute extinctions). Chronic disturbances, such as physicochemical habitat degradation, are pervasive. And in an uncertain future, these types of chronic disturbances are likely to pose significant challenges to Department of Defense (DoD) managers. The project objective was to address each of these two knowledge gaps—accounting for multiple types of species’ associations and predicting network responses to chronic disturbances—using North American freshwater fishes as a model system.
The project incorporated an exceptionally large, standardized, pre-existing database of fish co-occurrences (the combined Environmental Monitoring and Assessment Program, and National Rivers and Streams Assessment datasets) within three United States (U.S.) biogeographic regions: the Mid-Atlantic coast, Mississippi River basin, and Pacific Northwest. In each region, fish co-occurrence networks were built for sampling sites that had been objectively classified, using a standardized scoring system based on benthic macroinvertebrates, as least, moderately, or severely disturbed. Different metrics of network stability, including connectivity and modularity, were then used to gauge network responses to environmental disturbance (i.e., comparisons among least, moderately, and severely disturbed sites). Species’ extinction (i.e., node deletion) experiments were also used to simulate network responses to species’ extirpations and to detect secondary effects on the remaining species. Finally, a spatially-explicit framework to study and model species’ associations within dendritic river networks was explored by superimposing the spatially-implicit fish networks on spatially-explicit, digital stream network maps.
Network responses to chronic environmental disturbance were variable among biogeographic regions and in general, less clear than expected. Median node degree and closeness centrality did, however, change in a consistent, predictable manner with disturbance: degree decreased and closeness centrality increased with increasing disturbance. Together, these indicators suggest that fish co-occurrence networks become smaller (i.e., fewer species and fewer links among species) and more compact (i.e., shorter average distances among species’ nodes) as disturbance increases. Notably, the identities of species’ hubs (i.e., most highly connected species) did not change as the networks became smaller; decreasing degree values were attributable to deletions of peripheral species with low connectivity. In all networks, distinct modules of highly interconnected species were detected and the numbers of modules tended to increase with disturbance. Functional traits analyses showed that the morphological, ecological, and life history characteristics of species within modules were highly similar, suggesting that modules may be repetitive functional ‘motifs’. If these modules are functionally redundant, they may increase overall network stability; perturbations that disrupt part or all of a module may be compensated for by an independent module. Extinction simulations identified individual species that may be most vulnerable to primary and secondary extinctions, depending on the algorithm used to simulate disturbance (e.g., species least tolerant of pollution or species at sites closest to a major DoD facility).
This Limited Scope project was a unique opportunity to generate new understanding of ecological networks within stream ecosystems and to begin to develop a process for incorporating this knowledge in conservation planning. Overall, the researchers did not find definitive or easily generalizable results, but they did succeed in demonstrating an explicit process to quantify network structure and dynamics using widely available data on species’ co-occurrences. Lessons learned are germane to most U.S. streams, as the sample data were broadly representative of stream conditions and fishes found throughout the U.S., as well as other types of ecosystems where the requisite co-occurrence data can be obtained. Results should be of particular value to DoD managers as they prepare to contend with uncertain environmental conditions in the future, given that fishes comprise the largest taxonomic group of at-risk vertebrate species on DoD property and thousands of kilometers of stream and river habitat are in close or immediate proximity to one or more DoD facilities.