Combining a segmentation procedure and the BaRatin stationary model to estimate nonstationary rating curves and the associated uncertainties

Abstract Streamflow time series data are fundamental to hydrological science and water management applications, which are commonly derived from stage-discharge rating curves. The rating curves are usually nonstationary due to temporal characteristics of stream channel and human activities such as indiscriminate dig of sediment. This implies the need for an efficient method to estimate nonstationary rating curves from an adequate dataset. This study employed a shifting method to estimate the nonstationary rating curves and the associated uncertainties. The method includes two steps: a coupled algorithm PELT&CROPS was first performed to divide the gauging samples into homogeneous families by detecting the multiple change points in a time series of a residual indicator. Then, the rating curves were calibrated with the BaRatin stationary model using these homogeneous gaugings within each stationary period. The method was applied at Shijiao Hydrological Station in China. We found that: (1) Set the lower and upper limit of parameter CROPS (changepoints for a range of penalties) to 0 and Inf respectively, the whole range of segmentation can be calculated. (2) Parameter Minseglen (minimum segment length) is of great significance to the segmentation result. More changepoints will be detected as the value of Minseglen decreasing. An appropriate Minseglen value should be assigned to find the optimal segmentation. (3) The stage-discharge relationship in the low flow part is more sensitive to the variation of river topography. Ten segmentations are needed to produce an accurate discharge for low flow prediction, while two segmentations for calibration can produce an accurate discharge for high flow prediction. (4) Information redundancy exists in the vast amount of gaugings. The BaRatin model can reduce the use of calibration data by employing stage-discharge function with a precise exponent from the hydraulic knowledge at a Hydrological Station.