Commentary: IUCN classifications under uncertainty

Abstract

We comment on a recent article by Newton (Environ. Model. Softw. (2010), 25, 15–23), which proposed a method, based on a Bayesian belief networks, for classifying the threat status of species under the IUCN Red List Categories and Criteria, and compared this method to an earlier one that we had developed that is based on fuzzy logic. There are three types of differences between the results of the two methods, the most consequential of which is different threat status categories assigned to some species for which the input data were uncertain. We demonstrate that the results obtained using the fuzzy logic approach are consistent with IUCN Red List criteria and guidelines. The application of Bayesian Networks to the IUCN Red List criteria to assist uncertain risk assessments may yet have merit. However, in order to be consistent with IUCN Red List assessments, applications of Bayesian approaches to actual Red List assessments would need an explicit and objective method for assigning likelihoods based on uncertain data.

Introduction

In a recent article, Newton (2010) proposed a method based on Bayesian Networks for red listing under uncertainty, and compared his method to RAMAS Red List (Akçakaya and Ferson, 2001). RAMAS implements the fuzzy-logic approach we proposed (Akçakaya et al., 2000), which classifies species based on the IUCN Red List Categories and Criteria (IUCN, 2001) and allows uncertain data to be incorporated into the assessments. Based on a comparison of results for 16 species, Newton (2010) stated that the fuzzy logic approach gives “anomalous” results (p. 20, 22), suggested that it is not as reliable (p. 22) or transparent (p. 15) as the alternative, and implied that its performance has not been evaluated sufficiently (p. 22). This paper has led to some concern among people using the criteria for species with uncertain data, who then doubt the validity of the fuzzy logic method. Our goal is to demonstrate that the results obtained are consistent with IUCN criteria. We do not intend to review or correct Newton's approach in this paper; we simply aim to point out that, contrary to what Newton (2010) implied, the results obtained using fuzzy logic are neither anomalous nor incorrect with respect to IUCN Criteria and guidelines (IUCN, 2001, IUCN, 2011; Mace et al., 2008).

The comparison by Newton (2010) involved 16 species that were assessed by both methods. There were 3 types of differences between the results of the two methods: (i) in 2 of 16 cases, the methods assigned the species to a different threat category; (ii) in 5 cases, both methods assigned the species to the same threat category, but the range of plausible categories were different; (iii) in a few additional cases, the criteria under which the species were listed differed. We discuss these three types of differences in separate sections below.