Statistical principles for ecological status classification of Water Framework Directive monitoring data.

Abstract

Bias, precision and confidence of the classification framework are crucial elements for decisions to invest large sums to improve the ecological quality. In this study, the statistical principles for classification in relation to WFD are outlined and exemplified. Indicator adjustment to seasonal variation and other significant covariates reduces bias and improves precision. Precision is generally improved using annual means with seasonal adjustment instead of seasonal means. For classification I argue that the balance between costs of monitoring and reduction measures is only fully maintained by the fail-safe approach. The required monitoring efforts to ensure a precise classification are substantially higher than envisaged in WFD, for nutrients and phytoplankton measurements as high as 500 observations to characterise a water body. It must be ensured that sufficient monitoring data become available for classification, while indicator bias and precision is improved through modelling and further development of measurement techniques.