Using nitrate as a tracer to constrain age selection preferences in catchments with strong seasonality

Excess export of nitrate to streams has been an environmental issue attracting attention across the world. Lumped models based on StorAge Selection (SAS) functions have been used widely to investigate this export processes in terms of hydrological and chemical responses in catchment outflux - both in discharge (Q) and in evapotranspiration (ET). The validity of these models depends on the appropriate choice of water age selection preferences, i.e. the parameterization of the SAS functions of Q and ET, which are often calibrated against observed in-stream tracer data. This study attempted to assess the usefulness of in-stream nitrate (N-NO3) concentrations for resolving the SAS functions in terms of plausibility (being physically meaningful) and parameter uncertainty, in a small agricultural catchment in central Germany that experiences strong seasonality. A spatially implicit model was built to simulate the dynamics of nitrate export. The data sets of in-stream nitrate concentrations, nitrogen surplus in the soil, and the median age of Q were used to constrain the model. The best-fit (to all data sets) results indicate high plausibility of the calibrated, temporally dynamic age selection preferences for Q for most time of a year, however, not for ET. The uncertainty analysis using a Monte Carlo approach suggests that lower uncertainty is associated with the SAS functions for Q than that for ET. The uncertainties of the SAS functions can be reduced by 60% combining the three datasets instead of using only in-stream concentration data. This study demonstrated the usefulness of combined data sets for adequately determining dynamic age selection preferences for Q, but not for ET. Better constraining selection preferences will require better in-stream concentration data for the dry period and additional data sets for ET (e.g. age of ET).

Automated summary

Excess export of nitrate to streams is a global environmental concern, and lumped models based on StorAge Selection functions are widely used to study this process. This study assessed the usefulness of in-stream nitrate concentrations for resolving the SAS functions in a small agricultural catchment, finding high plausibility for Q but not for ET. Uncertainty analysis showed lower uncertainty for SAS functions for Q than ET, and data combination reduced uncertainties by 60%.