The problem of water body status misclassification-a Hierarchical Approach.

This article addresses the issue of estimating probability of misclassification (PoM), when assessing the status of a water body (w.b.). The standard deviation of a monitoring data is considered a good measure of the uncertainty of the assessed w.b. status. However, when PoM is to be estimated from the biological data, a problem caused by too few monitoring data emerges. The problem is overcome by developing Monte-Carlo models to simulate sufficient synthetic measurements of these elements, thereby accounting for random "disturbances" in the measurements. At each level of a procedure, called the Hierarchical Approach, values of PoM were derived from the Monte-Carlo-simulated data as for the assessment of w.b. status. It is assumed in the Hierarchical Approach that PoMs on each upper level can be estimated by processing PoMs inherited from the lower levels. Data from the river monitoring systems in three Polish regions were used in the study. Values of PoM calculated for biological elements show that 70-80% of cases belong to < 0.0, 0.1 > interval, whereas PoMs for physico-chemical elements in only 20% belong in this interval whereas for 25-40% of cases, PoMs are greater than 0.5. Moreover, when analyzing PoMs for cases when the w.b. status was classified as good, 22-52% of them are characterized by 0.5 or higher probability to be assessed wrongly. These pessimistic results suggest the need for formulation of new directions for future research in determining the PoM (in general, the uncertainty) of the w.b. status estimated from monitoring data.

Probability of misclassifying biological elements in surface waters.

Measurement uncertainties are inherent to assessment of biological indices of water bodies. The effect of these uncertainties on the probability of misclassification of ecological status is the subject of this paper. Four Monte-Carlo (M-C) models were applied to simulate the occurrence of random errors in the measurements of metrics corresponding to four biological elements of surface waters: macrophytes, phytoplankton, phytobenthos, and benthic macroinvertebrates. Long series of error-prone measurement values of these metrics, generated by M-C models, were used to identify cases in which values of any of the four biological indices lay outside of the "true" water body class, i.e., outside the class assigned from the actual physical measurements. Fraction of such cases in the M-C generated series was used to estimate the probability of misclassification. The method is particularly useful for estimating the probability of misclassification of the ecological status of surface water bodies in the case of short sequences of measurements of biological indices. The results of the Monte-Carlo simulations show a relatively high sensitivity of this probability to measurement errors of the river macrophyte index (MIR) and high robustness to measurement errors of the benthic macroinvertebrate index (MMI). The proposed method of using Monte-Carlo models to estimate the probability of misclassification has significant potential for assessing the uncertainty of water body status reported to the EC by the EU member countries according to WFD. The method can be readily applied also in risk assessment of water management decisions before adopting the status dependent corrective actions.