Modelling of trace metal transfer in a large river under different hydrological conditions (the Garonne River in southwest France)

Abstract

The modelling of trace metals (TM) in rivers is highly dependent on hydrodynamics, the transport of suspended particulate matter (SPM) and the partition between dissolved and particulate phases. A mechanistic, dynamic and distributed model is proposed that describes the fate of trace metals in rivers with respect to hydrodynamics, river morphology, erosion-sedimentation processes and sorption–desorption processes in order to identify the most meaningful parameters and processes involved at the reach scale of a large river. The hydraulic model is based on the 1-D Saint Venant equation integrating real transects to incorporate the river's morphology. The transport model of dissolved species and suspended sediments is based on advection–dispersion equations and is coupled to the one-dimensional transport with inflow and storage (OTIS) model, which takes transient storage zones into account. The erosion and sedimentation model uses Partheniades equations. Finally, the transfer of trace metals is simulated using two parameters, namely the partition coefficient (Kd) and the concentration of TM in the eroded material. The model was tested on the middle course of the Garonne River, southwest France, over an 80km section under two contrasting hydrological conditions (80m3s−1 and 800m3s−1) based on measurements (hydrology, suspended sediments, particulate and dissolved metals fractions) taken at 13 sampling stations and tributaries. The hydrodynamic model was calibrated with discharge data for the hydraulic model, tracer experiments for the dissolved transport model and SPM data for the erosion-sedimentation model. The TM model was tested on two trace metals: arsenic and lead. Arsenic was chosen for its large dissolved fraction, while lead was chosen for its very important particulate fraction, thus providing contrasting elements. The modelling of TM requires all four processes to be simulated simultaneously. The presented model requires the calibration of ten parameters divided in four submodels during two hydrological conditions (low and high flow). All parameters could be explained by the physical properties of the case study, suggesting that the model could be applied to other case studies. The strategy of using different datasets under different hydrological conditions highlights: (a) the importance of transient storage in the study case, (b) a detailed description of the erosion and sedimentation processes of SPM, and (c) the importance of TM eroded from the sediment as a secondary delayed source for surface water.