Causal networks as interpretive structures for multi-metric indices of ecological integrity

Background/Question/Methods

There is a large and growing emphasis on assessment and monitoring programs designed to provide decision makers with synopses of the health/condition of natural resources. A common approach uses multi-metric indices (MMIs), such as an Index of Biotic Integrity or Index of Ecological Integrity. Such systems typically seek to identify biotic or ecologic metrics with demonstrated responsiveness to human disturbance and combine those metrics into a MMI. The scientific enterprise of index construction is of major importance for the distillation of detailed measurements to inform decision makers. At the same time, the interpretability and utility of MMIs can be questioned. The objective of this study has been to apply MMI methods to wetlands Vital Signs monitoring data from Acadia National Park and Rocky Mountain National Park. Of primary interest has been a consideration of the scientific interpretability and management utility of such indices. To provide an interpretive structure for MMIs, we used the Vital Signs data to construct hypothesized causal networks relating elements of human disturbance to biotic and ecological responses. Evaluation of hypothesized networks using Bayesian structural equation modeling led to the development of Bayesian expert system networks to be used for probabilistic reasoning about the systems.

Results/Conclusions

Highlighted by the results from Acadia National Park wetlands, it can be shown that successful development of causal networks illuminating the linkages between human disturbance activities and biological/ecological responses can be a great aid to interpretation, knowledge application, and interventions. For Acadia wetlands, we conclude that the dominant impacts from human activities are hydrologic alterations and eutrophication. These wetlands appear to be quite sensitive to both impounding (lengthening of hydroperiods), which leads to a substantial loss of biodiversity in the system, and elevated mineral nutrients, which contributes to both a loss of nutrient-conserving Sphagnum spp. communities and invasion by aggressive dominating Typha spp. The predictive network developed reveals key sensitivities and illustrates ways that successful interventions can help preserve the integrity of these systems. The methodology presented represents a very general, integrated solution to causal network construction and network application under a Bayesian framework.