Wavelet and index methods for the identification of pool–rifflesequences

2018 
Abstract. The accuracy of hydraulic models depends on the quality of the bathymetric data they are based on, whatever the scale at which they are applied (e.g., 2D or 3D reach-scale modeling for local flood studies or 1D modeling for network-scale flood routing). The along-stream (longitudinal) and cross-sectional geometry of natural rivers is known to vary at the scale of the hydrographic network (e.g., generally decreasing slope, increasing width, etc.), allowing parameterizations of main cross-sectional parameters with proxy such as drainage area or a reference discharge quantile (an approach coined downstream hydraulic geometry, DHG). However, higher-frequency morphological variability is known to occur for many stream types, associated with varying flow conditions along a given reach: alternate bars, pool-riffle sequences, meanders, etc. To better take this high-frequency variability in bedforms into account in hydraulic models, a first step is to design robust methods to characterize the scales at which it occurs. In this paper, we propose and benchmark several methods to identify bedform sequences in pool-riffle morphology, for six small French rivers: the first one called the index method, based on three morphological and hydraulic descriptors; the second one called wavelet ridge extraction, performed on the continuous wavelet transform (CWT) of bed elevation. Finally, these new methods are compared with the bedform differencing technique (BDT, O’Neill and Abrahams (1984)), compared by computing a score that gives a percentage of agreement along the total surveyed length and by calculating the number of bedforms and the pool spacings for each method. The three methods were found to give similar results on average for wavelength estimation, with agreement from 64% to 84% and a similar number of bedforms identified. The filter-like behavior of the wavelet ridge analysis tends to give more robust results for the estimation of mean bedform amplitude, which varies from 0.30 to 0.81 with an SNR (signal-to-noise ratio) from 2.68 to 7.91. Otherwise, BDT gives higher mean bedform amplitude but lower SNR values from 0.85 to 1.73.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    0
    References
    0
    Citations
    NaN
    KQI
    []