ABSTRACT
A rating curve is an indirect method of estimating the discharge in rivers based on water level measurements. Continuous measurements of water levels are transformed to discharge thanks to the rating curve. A rating curve is traced for given hydraulics conditions and may vary over time as a result of natural processes such as erosion, sedimentation and seasonal vegetation growth which occur and change the river bed morphology. This implies that sometimes the rating curve is no longer sufficient and adequate to describe the real shape of the stage-discharge relationship. Direct discharge measurements (gaugings) are needed to build the stage-discharge relationship. The computed rating curve is the one that fits the best a serie of gaugings supposed to illustrate the existing hydraulic conditions at its time of establishment. As soon as a new gauging is made, the establishment of a new rating curve has to be performed if the gauging is quite far from the existing rating-curve (if a rating curve shift is detected). Most of the time, many gaugings are available in a hydrometric station, and the selection of the ones that were made in the same hydraulic conditions must be determined to build the new rating curve the most representative of the real hydraulic conditions at this time. In most unstable hydrometric stations where the river bed morphology varies often over time because of natural processes such as erosion and sedimentation, a new gauging is most of the time synonym of a new rating curve. Very often, the choice of the consistent gaugings to compute the most accurate rating-curve may become really tricky. Knowing the historical facts of a hydrometric station, and more precisely all the gaugings, this article proposes a methodology to devide the whole cloud of gaugings into chronological hydraulically homogeneous families by applying existing time series segmentations (Hubert et al. (1989), Kehagias and Fortin (2006)). Big events such as flood or landslides may be responsible for non negligible rating curve shifts. The aim of this study is to set a methodology proving that existing time series segmentation may be applied in the field of hydrometry to detect brutal hydraulic changes. Since streamflow data collected thanks to rating curves allow real time monitoring of rivers (hydro meteorological forecasts at points of interests), as well as hydrological studies and the sizing of structures, such a study will permit to ensure their quality.
