A multi-method analysis of bedload transport with seismometers and with hydrophones
Abstract:
IntroductionThe understanding of the short term variation in bedload flux is relevant for theoretical approaches - the scrutiny of bedload formulae is important for the planning of canal or river design. Extreme precipitation events may occur more frequently in the future as a consequence of climate change, and as a result the rate of bedload transport in rivers may considerably increase. At the current stage of development, bedload measuring devices have limited possibilities for a detailed investigation of mass transport kinetics with a temporal resolution shorter than one second. Therefore, we introduce a new experimental concept, a multi-method analysis of bedload transport with seismometers and impact-plate sensors (hydrophones).
Material and methods
Hydrophones, broadband seismometers, and sediment traps were installed at the measurement station in the small Koulbich gravel-bed brook in Northern Luxembourg (Colpach). Broadband seismometers are used to record of the ambient seismic field continuously (during different flood events as well as during baseflow). Measurements are performed with one sensor installed on the riverbank and two others installed further from the riverbed in order to discriminate signals originating only from the river. Our hydroacoustic set-up is constituted by a piezoelectric hydrophone acting as a bedload-induced vibration sensor in contact with a steel plate located at the bottom of the streambed (Krein et al., 2008). In addition, two ultrasonic level and Doppler flow meters record flow velocity and water level to estimate the turbulence degree. A hydro-morpho-dynamic model (Telemac3d-Sisyphe) has been set up along a 120 m section at the measurement site. Riverbed material samples were collected and sieved to estimate the riverbed material-grading curve (figure).
Evidence of hysteresis behaviour, estimation of bedload quantities and material characterization
The bedload sampler trapped the material after it moved over the hydrophone-plate system. Rocks and gravel-stones that produced the vibration signals were picked from the sampler, and their sizes and weights determined. The total power of single flood events calculated from the impact plate measurements is well correlated with the quantity of the transported material during the same flood event. The relation between suspended sediment concentration, bedload and water discharge is highly variable, and the use of sediment and bedload rating curves was considered unsuitable for predicting suspended sediment or bedload loads. Clockwise loops were the most common, which was attributed to the erosion or remobilization of sediment previously deposited on the channel bed or caused by mobilized bedload from the direct upstream section of the measurement site. During summer flood events the highest transport rates occur at the beginning of the rising limb. This is due to the rising transport capacity of the flow and the presence of loose, unconsolidated material. During winter flood events the bedload transport shows anticlockwise hysteresis. The sampled events were analysed to describe rainfall, runoff and sediment transport relationships. Better correlations were found between suspended sediment or bedload loads and runoff parameters (i.e. peak discharge and event runoff), than between loads and rainfall parameters (i.e. event precipitation, antecedent rainfall and maximum rainfall intensity). A parsimonious data-driven model (M5 modular trees) was used to simulate sediment loads in response to the identified controlling variables. Dominant antecedent hydro-meteorological conditions (e.g., antecedent precipitation depths, antecedent precipitation indices) act as the most significant controlling factors on the magnitude of sediment transport during episodic events.
Enhancing the characterization of bedload transport using the analogy between the Hertz contact theory and impact-plate measurements
In view of the high signal-to-noise ratio of the recorded impact signals, we aim to improve the classical impact counting or total power processing in order to derive more information on time-varying bedload properties (Rickenmann et al., 2012). According to the Hertz contact theory and acoustic emission method literature, grain size can be estimated from acoustic measurements. Hertzs theory allows calculating the force time history that round ball imposes on an elastic plate after being dropped at normal incidence. The interaction of stones with the metal plate is much more complex with different motions (saltation, rolling, sliding) and shapes (round, flat). However, a certain analogy with the Hertzian contact might be envisaged since the force pulse is highly dependent on the ball size and mass. The impact signal generated by coarse or fine sediments should statistically have a different signature, despite the high variability of grain shapes, impact angles and velocity. When a sediment particle impacts the plate, the amplitude and the frequency of the first arrival waveform are the two fundamental properties related to the force that the bedload imposes on the plate and the contact time defined as the duration when the impact force is nonzero. Therefore, a relationship should exist between the frequency/amplitude attributes and the size of sediment and could be highlighted by using an appropriate signal processing technique. Nevertheless, recorded signals after impacts exhibit complex waveforms due to boundary reflections and rebounds. To isolate and characterize the first arrival, we developed a complete and advanced processing algorithm, based on a signal decomposition technique (Bardainne et al., 2006), consisting in two main steps:
- First break picking of each coherent first arrival in the time domain using a detection method similar to STA/LTA (short-term average to long-term average) technique in seismology, which permits differentiating a signal onset within a noisy time history.
- Full characterization of the first waveform based on a high dimensional decomposition method (the chirplet atomic decomposition) to get an accurate estimation of the frequency content and amplitude. This method provides an optimal reconstruction of the selected waveform in terms of chirp atom characterized by 7 parameters which are a generalization of 2-D information obtained by wavelet transform in the time-frequency domain.
We show through laboratory flume experiments that the two attributes aforementioned are valuable criteria to differentiate many grain size classes (Barriere et al., 2014). This method will be soon applied to field measurements and compared to the grain size distribution of collected sediments.
Does the bedload have a seismic signature in small gravel-bed rivers?
Bedload surrogate monitoring based on seismological observations have been envisaged in the last years. This so-called fluvial seismology has been developed in the context of sediments transport in mountain rivers or debris flows (Burtin et al., 2010; Tsai et al., 2012). We propose here to study less energetic streams flowing in the low mountain range to investigate the seismic detection threshold of sediments motion. The main results were obtained for a typical long-lasting winter flood event (October 10th to November 14th2013) with water level variations from 0.17m to 0.6m. After selecting a non-human disturbed time window, a significant increase of the seismic noise level in conjunction with an increase of the water level was observed for the seismometer close to the river whereas no similar and significant trends were noticed for the two others seismometers. By analysing in detail the variations of frequency content during this highest water discharge (double-peak event of similar magnitude), we would like to discriminate the signature of the water turbulences and the inception of motion of coarse sediments potentially detectable by the seismometer. During the first peak flow, the relative level of seismic noise evolves as a function of the water level. During the second peak, the correlation between seismic noise and water level still exists but the frequency content of seismic signals exhibits significant differences between the two peak flows. The dominant frequencies higher than 30 Hz for the first peak shift toward lower frequencies during the second one. The level of seismic noise is also clearly stronger for the second peak (+ 10/15 dB) whereas the water level is similar for the two peaks. This observation could be a seismic signature of the bedload transport but further investigations are needed to better understand the source of river-induced seismic noise. A joint analysis of such seismic results with a 3-dimensional hydro-morphodynamic simulation is an ongoing study. This seismic approach is now complemented by impact plate measurements to constraint the potential bedload-related seismic measurements.
Conclusion
The potential advantages of acoustic/seismic bedload surrogate monitoring are a high temporal resolution in real time, the possibility to register the bedload transport during high turbidity and the observation under undisturbed conditions. We can show that the total power of impact signals is consistent with the bedload rates and that impact-plate measurements can describe the highly time-resolved kinetics of transported material. Moreover, the new processing approach developed here offers the opportunity to better understand the bedload signature of vibration measurements and is promising in view of quantifying grain size distribution during bedload transport. Our current and further studies aim to apply this method, validated by flume experiments, to impact measurements in the Koulbich brook. Finally, another ongoing study consists in testing an attractive non-invasive technique called fluvial seismology by performing a joint analysis of seismic measurements, collected bedload sediments, impact plate data and sediments transport simulation. The main interests are to investigate the threshold of seismic detection of sediments motion in streams and assess the validity of this seismic approach to constrain hydrological transport models. First results suggest the potential of monitoring seismically a wider range of rivers than previously investigated.
References
Bardainne, T., Gaillot, P., Dubos-Sallee, N., Blanco, J. & Senechal, G. (2006). Characterization of seismic waveforms and classification of seismic events using chirplet atomic decomposition. Example from the Lacq gas field (Western Pyrenees, France). Geophysical Journal International 166, 699-718.
Barriere, J., Oth, A., Schenkluhn, R., Krein, A. (2014). Enhancing the characterization of bedload transport using the analogy between the Hertz contact theory and plate hydrophone measurements. Oral presentation, EGU General Assembly 2014, Vienna, abstract EGU2014-4134
Burtin, A., Vergne, J., Rivera, L., & Dubernet, P. (2010) Location of river-induced seismic signal from noise correlation functions. Geophysical Journal International 182, 1167-1173.
Krein, A., Klinck, H., Eiden, M., Symader, W., Bierl, R., Hoffmann, L. & Pfister, L. (2008). Investigating the transport dynamics and the properties of bedload material with a hydro-acoustic measuring system. Earth Surface Processes and Landforms 33, 152-163.
Rickenmann, D., Turowski, J.M., Fritschi, B., Klaiber, A. & Ludwig, A. (2012). Bedload transport measurements at the Erlenbach stream with geophones and automated basket samplers. Earth Surface Processes and Landforms 37, 1000-1011.
Tsai, V.C., Minchew, B., Lamb, M.P., & Ampuero, J-P. (2012) A physical model for seismic noise generation from sediment transport in rivers. Geophysical Research Letters, 39, L02404.
Acknowledgements
