Spatial and temporal characterisation of "Stream Water Bodies" quality

This research aimed to understand how to interpolate discrete measurements, in space and time, in order to calculate physico-chemical indicators in rivers, which are required by the European Water Framework Directive. Linked to this issue, several questions were addressed. Are the different methods used to calculate temporal 90th-percentiles at a given site equivalent? How does this legal indicator vary in space? Does the French National Basin Network provide enough information to make consistent water quality characterization? The daily outputs of the ProSe model applied to the Seine River were used as proxies to compare different calculation methods of the 90th-percentile. The results deduced from the exhaustive model were compared to those calculated, after sampling the outputs according to the monitoring network sampling scheme. Two calculations of the temporal 90th-percentile at a given site were examined: the classical method based on the empirical percentile function and a slightly more complex method that includes temporal weighting and linearization of the empirical percentile function. This second method reduced the estimation bias of the 90th-percentile induced by irregular and/or few measurements. Three methods for spatializing the 90th-percentiles were tested to obtain occurrence percentages of the percentiles for each quality class in each "Stream Water Body": the "failure principle" consists in keeping only the worst site; the second approach calculates the proportion of sites located in each quality class; the third method allocates an influence segment to each measurement site. Spatializing temporal percentiles in "Stream Water Bodies" by influence segments led to a marked improvement in occurrence percentage estimations and revealed the need to take into account the spatial configuration of measurement sites when calculating a quality indicator.