Quantifying measurement uncertainty in real time continuous water quality monitoring: methodological approaches and...

A long-term environmental programme (OPE) was set up in 2007 by the French National Radioactive Waste Management Agency as a research tool observing simultaneously various environmental compartments in the north-eastern part of France. Since 2012, six continuous water quality monitoring stations created by the OPE assess the variability in time and space of the water quality of three rivers flowing in limestone catchments with distinct impacts of groundwater/surface interactions. We developed a novel approach to estimate sensor measurement uncertainty at the six OPE river stations by calculating the 90th percentile of differences between monthly quality controls and in situ sensor readings. Performances under controlled and real conditions were also compared, showing that this method based on percentiles 90 using ongoing quality controls is more realistic than uncertainty evaluations carried out under controlled conditions. This is partly because it takes into account real conditions over a hydrological year (temperature, flow of watercourses, interferences, etc.) as well as long-term deployment conditions (biofouling). The measurement uncertainty in real time is a prerequisite to avoid over-interpreting variations that are within the measurement uncertainty and thus identify significant trends over time. After accounting for uncertainty, the variability of conductivity, dissolved oxygen, and pH measured using sensor measurements in these rivers during 14 years could be analysed confidently. This in turn allowed us to interpret differences in the impacts of groundwater-surface interactions on river composition and confirm the extent and variability of dissolved oxygen depletion in these rivers.