River discharge estimation from radar altimetry: Assessment of satellite performance, river scales and methods

•Sentinel-3 combination with CryoSat-2 and AltiKa gives the most accurate water height retrievals.•The rating curve method outperforms hydraulic- and model-based methods of discharge estimation.•River depth for Manning model can be estimated from regional depth = f(width,deposits) relations.•The accuracy of altimetric discharge retrievals decreases in parallel with the river size. The ENVISAT, Jason-2 and -3, Sentinel-3A, CryoSat-2, and AltiKa satellite altimeters were used to estimate the discharge of two Arctic rivers: the Ob and Pur in western Siberia. The accuracy of the water height estimates from ENVISAT varied with river size from 0.63 m for the Ob (~2 km in width) to 1.1 m for the Pur (~0.5 km in width). A new method for water level estimation combining the CryoSat-2, AltiKa, and Sentinel-3A altimeters provided better height accuracy (0.49 m for the Ob). The rating curve method of discharge estimation outperforms the Manning formulation, on the Ob, with a root mean square error (RMSE) of 13% vs 20% for daily discharge and 1% vs 5% for annual discharge. Sensitivity analysis performed on the Manning formulation showed high sensitivity to the river depth parameterisation. A functional relationship was identified between the depth of rivers, in the north of western Siberia, and their widths and deposits types. The combination, for the Ob, of the rating curve method with the multi-satellite water level retrieval algorithm provided higher discharge accuracy than previous studies relying on satellite measurements, on numerical modelling or on their combination. A synthesis of worldwide river altimetry studies supported our finding that the accuracy of altimetric discharge estimations decreases for narrower rivers. The rating curve method applied to Jason 2/3 height measurements allows estimation of daily discharges with RMSE of 18% for the wide Ob River but 38% for the narrow Pur River.