Uncertainties in data and models to describe event dynamics of agricultural sediment and phosphorus transfer

Mathematical models help to quantify agricultural sediment and phosphorus transfers and to simulate mitigation of pollution. This paper develops empirical models of the dominant sediment and phosphorus event dynamics observed at high resolution in a drained and undrained, intensive grassland field-scale lysimeter (1 ha) experiment. The uncertainties in model development and simulation are addressed using Generalized Likelihood Uncertainty Estimation. A comparison of suspended solids (SS) and total phosphorus (TP) samples with a limited number of manual repeats indicates larger data variability at low flows. Quantitative uncertainty estimates for discharge (Q) are available from another study. Suspended solids-discharge (SS-Q) hysteresis is analyzed for four events and two drained and two undrained fields. Hysteresis loops differ spatially and temporally, and exhaustion is apparent between sequential hydrograph peaks. A coherent empirical model framework for hysteresis, where SS is a function of Q and rate of change of Q, is proposed. This is evaluated taking the Q uncertainty into account, which can contribute substantially to the overall uncertainty of model simulations. The model simulates small hysteresis loops well but fails to simulate exhaustion of SS sources and flushing at the onset of events. Analysis of the TP-SS relationship reveals that most of the variability occurs at low flows, and a power-law relationship can explain the dominant behavior at higher flows, which is consistent across events, fields, and pathways. The need for further field experiments to test hypotheses of sediment mobilization and to quantify data uncertainties is identified.