Monte-Carlo methods to assess the uncertainty related to the use of predictive multimetric indices

Abstract

The publication of the Water Framework Directive by the European commission in 2000 has promoted the development of many multimetric biological indices to assess the ecological status of European waterbodies. These ecological assessments are based on the measurement of deviations between a metric’s (characteristic of assemblages) observed values (obtained by sampling) and a metric’s expected values in the absence of anthropogenic stressors (reference conditions). In addition, the confidence in the ecological status evaluation provided by the different biological indices is required. Numerous sources of uncertainty due to sampling variability or operator bias, for example, are often considered on observed metric values, whereas uncertainty associated with expected metric values are seldom discussed. In this study, we developed a methodology based on Monte-Carlo methods to assess the uncertainty associated with the establishment of reference values for multimetric predictive indices. This was done by randomly generating reference values and propagating the uncertainty throughout the computation of the index. This methodology can be applied to a wide variety of indices as long as it is possible to make assumptions about the statistical distributions of some of the index’s numerical components (e.g. coefficients of the statistical models, metric values). The European Lake Fish Index was used to illustrate the methodology and show how this method can provide valuable information on the confidence in the ecological status defined by the index. These results also revealed that the degree of uncertainty varied between the ecological classes, which were highest for the “Moderate” class and lowest for the “Poor” and “High” classes for the ELFI.