Using process-based models to filter out natural variability in observed concentrations of nitrogen and phosphorus in river water.

Abstract

Advances in process-based modelling of loads of nitrogen and phosphorus carried by rivers have created new possibilities to interpret time series of water quality data. We examined how model runs with constant anthropogenic forcing can be used to estimate and filter out weather-driven variation in observational data and, thereby, draw attention to other features of such data. An assessment of measured and modelled nutrient concentrations at the outlets of 45 Swedish rivers provided promising results for total nitrogen. In particular, joint analyses of observational data and outputs from the catchment model S-HYPE strengthened the evidence that downward trends in nitrogen were due to mitigation measures in agriculture. Evaluation of modelled and observed total phosphorus concentrations revealed considerable bias in the collection or chemical analysis of water samples and also identified weaknesses in the model outputs. Together, our results highlight the need for more efficient two-way communication between environmental modelling and monitoring.