Large scale nutrient modelling using globally available datasets: A test for the Rhine basin

Abstract

Summary Nutrient discharge to coastal waters from rivers draining populated areas can cause vast algal blooms. Changing conditions in the drainage basin, like land use change, or climate induced changes in hydrology, may alter riverine nitrogen (N) and phosphorus (P) fluxes and further increase the pressure on coastal water quality. Several large scale models have been employed to quantify riverine nutrient fluxes on a yearly to decadal timescale. Seasonal variation of these fluxes, governed by internal nutrient transformations and attenuation, is often larger than the inter-annual variation and may contain crucial information on nutrient transfer through river basins and should therefore not be overlooked. In the last decade the increasing availability of global datasets at fine resolutions has enabled the modelling of multiple basins using a coherent dataset. Furthermore, the use of global datasets will aid to global change impact assessment. We developed a new model, RiNUX, to adequately simulate present and future river nutrient loads in large river basins. The RiNUX model captures the intra-annual variation at the basin scale in order to provide more accurate estimates of future nutrient loads in response to global change. With an incorporated dynamic sediment flux model, the particulate nutrient loads can be assessed. It is concluded that the RiNUX model provides a powerful, spatial and temporal explicit tool to estimate intra-annual variations in riverine nutrient loads in large river basins. The model was calibrated using the detailed RHIN dataset and its overall efficiency was tested using a coarser dataset GLOB for the Rhine basin. Using the RHIN dataset seasonal variable nutrient load at the river outlet can be satisfactorily modelled for both total N (E = 0.50) and total P (E = 0.47). The largest prediction errors occur in estimating high TN loads. When using the GLOB dataset, the model efficiency is lower for TN (E = 0.12), due to overestimated nutrient emissions. For TP, the model efficiency is only slightly lower (E = 0.36) in comparison to the RHIN dataset. Despite the lower model efficiencies for the GLOB dataset, we conclude that this dataset provided reasonably good estimates of seasonal nutrient loads in the Rhine basin and is considered promising for application to other, less documented, large river basins.