Eye of Newt

Abstract

Whoever would identify a wetland objectively must cook up some combination of evidence from hydrology, soil, and vegetation. Many recipes have been proposed, but all have proven unappealing to many and downright nauseating to some. Some favor one ingredient over the other two, while others insist on all three. These differences go beyond mere matters of taste; they relate directly to probability of error and feasibility of practice. Use of observations or measurements to make conclusions always involves some probability of error. Statistics is the discipline to which we turn in our attempt to attach probabilities of error to a particular judgment. That judgments about wetlands lie within the reach of statistics seems to be forgotten most of the time. Statistics, while regarded by many as an ugly discipline, has been reborn in several more comely forms. One of these is risk analysis, which involves the assignment of probability to outcomes that society views as undesirable. Given that risk analysis has been recently as much a rage as statistics itself was a generation ago, wetland identification should be steeped in it, but this is not the case. In arguing over the identification of wetlands, critics of a particular type of evidence are likely to say that it is unreliable. No such generalization is reasonable; the reliability of a particular type of evidence depends on the situation. For example, a plant community strongly dominated by obligate wetland plant species (or equivalently, showing a prevalence index below 2) will support the identification of a wetland with very little risk of error if there has been no recent change in hydrology. The risk thus could be framed as a statement of conditional probability: given that hydrologic conditions have not changed recently, strong dominance of the plant community by obligate wetland species indicates the presence of wetland with a probability exceeding 90%.2 On the other hand, dominance of facultative species (e.g., prevalence index of 2.5) would be a very different matter.