By applying ERA5 monthly temperature data, periods of likely ice presence were identified and removed, resulting in a more consistent rating curve. This filtering allows the construction of a more representative and robust rating curve, avoiding the influence of anomalous hydraulic states induced by ice, while maintaining a unified Bayesian framework. Figure 4: Rating curve for the Near-Brookings station (Mississippi basin) before (left) and after (right) excluding data points associated with temperature variations (below 0°C).
4.2.3.3 Case 3 – Artic rivers In Arctic regions, the relationship between water surface elevation (WSE) and discharge is often non- uniform due to seasonal ice cover, ice breakup, and complex flow dynamics near river confluences. In such contexts, a single rating curve may be inadequate. Instead, a set of rating curves, each tailored to a specific hydrological regime—such as flood rise, recession, or ice-covered periods—is employed. This approach requires prior knowledge of the river’s seasonal behavior and the applicability range of each curve.
The impact of river ice on flow hydraulics can be detected using remote sensing and altimetry observations acquired simultaneously with WSE retrievals (Zakharova et al., 2021). Previous studies have demonstrated that using distinct rating curves for different flow conditions improves discharge estimation accuracy (Zakharova et al., 2020). Classical power-law equations are used for flood rise and recession periods, while low-degree polynomials may better capture the hydraulic behavior during the ice-covered period.
For example, on the Ob River, an automated method based on altimetry was tested to detect ice-on and ice-off periods from 2008 to 2019, which enabled the calibration and validation of a winter-specific rating curve. For years not covered by this product, optical imagery from Landsat and MODIS was used. The spring flood rise subset was extracted directly from the WSE time series. On four Arctic test sites—located on the Colville, Mackenzie, and Lena rivers—two rating curves were built using a modified Bayesian method: one for flood recession, and another combining winter and flood rise periods.
In this modified method, only the parameters “a” and “z ₀ ” were estimated probabilistically, while the exponent “b” was fixed to allow expert adjustment of the curve shape for each hydrological regime. This regime-based approach enhances discharge estimation in Arctic environments by accounting for seasonal variability in flow hydraulics and by leveraging multi-source satellite data. D.3 River Discharge ATBD. Reference: RD_CCI-0006-ATBD - Issue 2.0 – 31/07/2025 Open/Public © 2019 CLS. All rights reserved. Proprietary and Confidential. 17/31