Accounting for discharge simulation performances and observation quality for modelling sediment at global scale

Water quality modelling is very often constrained beforehand by the performance of discharge modelling. When the model is set up at global scale, this discharge performance usually greatly varies in space and can therefore be a limiting factor in many regions around the globe. Besides discharge performances, the quality of the observations themselves can also be highly heterogeneous in space and among datasets. Modeller then has to find a compromise between being restrictive on the quality of his dataset or having a good spatio-temporal representativeness of the various hydrological conditions at global scale. This often relies on subjective thresholds.

This work proposed a more objective calibration strategy that aims to consider both aspects explicitly: observation quality and model performance on discharge. It leads to the construction of two scores that are assigned to each water quality station quantifying their reliability for model calibration and evaluation. The average of those two scores is then used as a weight in the objective function to emphasise the training on the most reliable stations.

The strategy is implemented for sediment modelling using the WW-HYPE model (Arheimer et al., 2019) at global scale. The score on discharge simulation performance is based on the regionalisation of the Nash-Sutcliffe Efficiency that is spatially interpolated at every sediment monitoring stations using inverse Ghosh distance weighting (de Lavenne et al., 2016). The score on the observation quality is based on the location of the station with respect to the catchment outlet and on the amount of data. A multi-objective calibration is performed to optimise parameters on two global databases, one on long terms sediment loads (730 stations) and one on sediment concentrations time series (1440 stations). The sensitivity of the model to this calibration strategy is analysed according to model performances and model outputs, such as sediment loads at global scale, in order to discuss the importance of considering this heterogeneity of the reliability of monitoring stations.

References

Arheimer, B., Pimentel, R., Isberg, K., Crochemore, L., Andersson, J. C. M., Hasan, A., and Pineda, L. (2019), Global catchment modelling using World-Wide HYPE (WWH), open data and stepwise parameter estimation, Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2019-111.

de Lavenne, A., J. O. Skøien, C. Cudennec, F. Curie, and F. Moatar (2016), Transferring measured discharge time series: Large-scale comparison of Top-kriging to geomorphology-based inverse modeling, Water Resour. Res., 52, 5555–5576, doi:10.1002/2016WR018716.