Expert Article Library

Fish, Wildlife, and Industrial Activities in Litigation

There are many plant and animal species considered to be threatened, endangered, or of special concern to regulators and the public. Correctly estimating population sizes, relationship to habitats, and potential effects of industrial activities is crucial to many lawsuits. Other environmental issues are more broad, such as quantifying relationships of industrial activities and natural ecosystems. The most effective approach to addressing these issues is to quantify causality and explain it in language understood by non-technical finders of fact.

Many environmental lawsuits involve concerns about "healthy ecosystems" or "sustainable populations." These are subjective concepts and cannot be directly measured, but are at the core of environmental permitting decisions by regulators and the public and are often the basis for legal challenges.

There are many subjective concepts we all use in everyday language. They cannot be directly measured and mean different things to each individual. Yet, we can indirectly put relative measures to them. Examples of linguistic variables are "tall," "heavy," "expensive," and "fast." We assign relative meaning to these by measuring height for tall, weight for heavy, cost for expensive, and speed for fast.

Abstract concepts such as "healthy ecosystems" and "sustainable populations" are not as easily measured. Social sciences (e.g., psychology, economics,
sociology) are based on abstract concepts like these and decades ago their practitioners adopted suitable statistical models to analyze them and apply results to real issues. Consider the concept if IQ. We assume that each individual has innate abilities over a broad range of actions but we cannot directly measure the aggregate of all these different abilities (the IQ score). Social scientists defined a set of causal variables that by convention represent overall IQ when each variable's relative contribution to the whole is determined. Environmental science is only now starting to use the same approach to address similar difficult concepts in natural ecosystems.

As an example, it is important to objectively define a sustainable population of Greater sage-grouse; natural resource operators, the public, and environmental NGOs all need to know. Potential causal variables might include habitat quantity, quality, and distribution, predation (nest/egg, juvenile, adult), fecundity (fertility of the population), demographics (births minus deaths), and anthropomorphic activities. These are measurable indicators hypothesized to contribute to sustainability. They may not all be independent (a prerequisite of frequentist statistical hypothesis testing models) so the analysis must include correlations. For example, habitat quality and distribution might be correlated and habitat quantity and nest predation might be correlated. There may also be many unobserved (latent) indicators that contribute to sustainability. Latent variables include measurement errors (did the surveyors actually see all the birds that were present during the site visit?), and factors we did not include because we were not aware of them.

These causal and response factors are drawn as a map with connections between them. This map (or directed graph) reflects the paths we hypothesize that explain the data we observe in our surveys of sage-grouse, their habitats, and potential predators. Path analysis is the first step to defining a sustainable population. The next step is analyzing each path's relative contribution to the population's size using a structural equation model (SEM).

The major difference between the familiar null hypothesis of regression models used to predict cause-and-effect and the SEM causal hypothesis is that the former requires the data to be fit to it and the null is rejected as an explanation of the observed data at the 0.05 probability level. The SEM hypothesis represents the expected values of the directed graph's path coefficients. If we are not satisfied with how closely the expected values match the observed values we can modify the graph (or the paths) until we achieve an acceptable fit of observed to expected values. The most common approach is to determine the hypothetical model with the maximum likelihood of explaining the observed data.

Fitting the accepted model of expected values to a different set of observed values confirms its generality. Using the validated model allows us to determine which indicator variables are most responsible for determining the population size so we know which need the greatest attention to maintain and increase population size. This statistical method is a technically sound and legally defensible approach to defining sustainable populations and can be communicated visually as well as in words to finders of fact.

Litigation involving abstract environmental and natural resource concepts can benefit by objectively quantifying causality. Path analysis and structural equation modeling can be applied to quantifying healthy ecosystems, sustainable populations, and other abstract concepts.