Applications of Causal Analysis: A Tiered Approach With Emphasis on Data Reliability
On this page:
Over the past decade, causal analysis has been increasingly used to diagnose environmental impairments within water bodies. Much of this work is guided by the U.S. EPA CADDIS program. We have adapted these tools to a variety of environmental issues involving observed or presumed stressors on terrestrial environments, air, coastal systems, deserts, and wetlands. Our analyses have examined biological, chemical, and physical stressors in these domains. In structuring these analyses, we find that a tiered approach offers a means of focusing the analysis on key questions. Each analysis level informs the need for further investigation sufficient to understand which individual or combination of stressors is causing an outcome. The method we developed has been used to support evaluations designed to achieve appropriate interventions and have also been presented in court proceedings including before the U.S. Supreme Court and International Court of Justice. We will use case examples to illustrate this tiered assessment strategy. Because causal analysis often involves a retrospective that uses extant information, we have become increasingly aware of the heightened need for rigorous data quality assurance methods. Extant lines of evidence supporting or refuting candidate causes can be of variable quality and quantity. These statistical and design variations among data sets can lead to misrepresentation of the relative importance of a candidate cause. A candidate cause represented by poor data may appear less important to the analyst than a cause for which there is considerable data. This is an inequity data quality issue that must be recognized when conducting causal analysis. We present an example along with guidance on how to ensure adequate data quality for the process. We will illustrate this with case studies including biological effects associated with sediment contamination, water quality effects, declines of bees, and an oil spill.