Resilience to disturbance is vital to the sustainability of ecological systems. However, determination of system resilience in the field is less than straightforward because of its dependence on the nature of the disturbance and on the choice of endpoints for system recovery. More frequently, we base assessments of ecological system health on measures of function, such as abundance, productivity, and biomass, and measures of structure, such as species richness and diversity, which are relatively simple to obtain in the field.

I conducted 14 experimental fish defaunations in 100-meter reaches of second- to fourth-order warmwater streams in the Red Cedar River watershed of lower Michigan. Stream habitat was left intact. I documented fish assemblage recovery in these reaches over the course of a month in order to describe the structure of recovering assemblages over time and to identify habitat or assemblage characteristics that influenced resilience. Fish assemblages recovered remarkably quickly, exhibiting high similarity to pre-disturbance structure within two days of defaunation. Pool habitat appeared to encourage recolonization, while pre-disturbance fish abundance was negatively related to abundance recovery. No relationship between resilience and multimetric indices commonly used to assess stream health, such as overall habitat quality or biotic integrity, was evident. Such metrics, although efficient in terms of cost and time, may not produce a complete picture of stream health.

The results of these experiments permitted an examination of the relationship between assemblage structure, function, and resilience as components of ecological health. There was no relationship between the pre-disturbance vigor or organization of the fish community and its resilience to disturbance. While I found that measuring post-disturbance changes in structure and abundance over time was useful in determining recovery, these resilience values are applicable only to one type of disturbance at one spatial and temporal scale. Although vigor and organization may serve as partial indicators of community health, a general model of the relationship of these variables to resilience would be invaluable for ecosystem conservation and management.

Finally, the structure of these defaunation experiments provided an opportunity to investigate habitat selection in a common stream fish, the creek chub (Semotilus atromaculatus Mitchill). Traditional methods of assessing habitat suitability rely on correlating habitat characteristics with the number of individuals occupying the habitat under natural conditions. Distribution is a behavioral phenomenon, however, and the majority of these correlations do not consider movement or density-dependent interactions that may influence the suitability of a particular habitat, and therefore, the dynamics of creek chub population recovery following to disturbance. In this study, I considered both the influence of physical habitat and density of conspecifics on distribution. Creek chubs preferred pools over runs, and runs over riffles. Alternative methods of measuring selection, including selection indices and immigration rates, revealed similar insights. Density of creek chubs in the study reaches, which ranged from zero immediately following defaunation to above pre-defaunation (natural) densities, had no apparent effect on habitat preferences.