Detection of Stage‐Discharge Rating Shifts Using Gaugings: A Recursive Segmentation Procedure Accounting for Observational and Model Uncertainties

Abstract

Introduction Rating CurvesRiver discharge, or streamflow, is one of the most important variables in hydrology and hydraulics. Hydrometric data are essential for the calibration of hydrological models, flood forecasting and warning, engineering design (of dam and bridges for example), and policy decisions related to water resources and environmental management. However, streamflow time series are not direct observations as streamflow cannot generally be continuously measured in natural rivers. Instead, the water level (also called "stage") is continuously monitored. Streamflow time series is hence derived from rating curves (Rantz, 1982; WMO, 2010), hereafter called "RCs", which are models transforming an input stage into an output discharge. These models are estimated using occasional stage-discharge measurements (also known as gaugings) and hydraulic constraints. The physical relation between discharge and stage is determined by hydraulic controls, that is, physical characteristics of the river section or channel influencing the flow: geometry, friction, longitudinal slope, head losses, etc. A hydraulic analysis of the site, through field expertize or modeling, allows identifying the succession or addition of hydraulic controls as flow increases (Le Coz et al., 2014).RCs are affected by many sources of uncertainty, including structural uncertainty (imperfection of the RC model), gaugings measurement uncertainty, and parametric uncertainty (estimation of RC parameters).

Streamflow monitoring at high temporal resolution based on non-contact instruments in a river prone to bathymetric shifts

“…The temporal limits of these periods were extracted from the time-stamping of the couples of points (H rad , V rad ). It is worth mentioning that, even in the absence of the V rad time series as it is the case for many hydrometric stations, it would have been possible to identifiy the bathymetric changes using an automatic shift detection algorithm such as the method presented recently by Darienzo et al (2021).…”

“…In fact, there is still some dispersion in the H rad -V rad relations for the different separated periods. A more detailed identification, still based on the analysis of the relations between V rad and H rad or based on the application of an automatic shift detection method (Darienzo et al, 2021) would be possible but it would lead to more sub-periods and we would face to the problem of lack of bathymetric data. The proposed breakdown is therefore a compromise between the identification of major bathymetric shifts and the availability of topographic data.…”

Assessment of uncertainties on stage-discharge rating curves: A large scale application to Québec hydrometric network

“…RC were first estimated using all gaugings at a station. Resulting RC are hereafter called baseline RC, as in Darienzo et al (2021). The analysis is performed when six gaugings or more are available, which is the case for the 173 stations under study.…”

“…The total number of GS finally obtained at a station corresponds to the number of periods and RC necessary to represent the stage-discharge relationship over the gauging period at this station. A similar approach was proposed by Darienzo et al (2021). These authors use a more complex approach based on multichange point Bayesian estimation and applied the proposed approach to a station on the Ardèche river (France).…”

Design flood assessment on the upper Rhine using historical data and accounting for discharge uncertainty

“…Using the correction n°1, we apply a segmentation procedure (Darienzo et al, 2021) on the annual maximum floods from 1808 to 2017. This method accounts for observational discharge uncertainties and expresses change points in terms of time (rather than position).…”