Debris flow seismic monitoring and warning

Standardization of measurement procedures and performances are important goals in every field of science and are in general intensely pursued by scientists. Certain natural phenomena, however, present particularly difficult challenges in this regard and many efforts are still needed to actually reach standardization and systematic performance of measurements. Debris flows certainly belong to this latter category. Due to their low frequency of occurrence, their short duration and their sudden and abrupt nature they are extremely difficult to be measured. Only instrumented basins where debris flows occur with a sufficiently high frequency per year allow systematic monitoring activities. Even though during the last decades several such basin have been instrumented, field measurement data are still scanty and methods of measurement are not yet sufficiently standardized. One of the goals of the European Territorial Cooperation project SedAlp “Sediment management in Alpine basins: integrating sediment continuum, risk mitigation and hydropower” (Alpine Space Programme 2007-2013) is to make some advancement in this direction. One of the expected outputs of the SedAlp project is a protocol on debris flow monitoring. Ground vibration sensors have been increasingly used and tested, during the last few years, as devices to monitor debris flows and they have also been proposed as one of the more reliable devices for the design of debris flow warning systems. The need to process the output of ground vibration sensors, to diminish the amount of data to be recorded, is usually due to the reduced storing capabilities and the limited power supply, normally provided by solar panels, available in the high mountain environment. There are different methods that can be found in literature to process the ground vibration signal produced by debris flows. In this paper we will discuss the two most commonly employed: the method of amplitude and the method of impulses. These two methods of data processing are analyzed describing their origin and their use, presenting examples of applications and their main advantages and shortcomings. The two methods are then applied to process the ground vibration raw data produced by a debris flow occurred in the Rebaixader Torrent (Spanish Pyrenees) in 2012. The detection of debris flows through seismic devices occurs at a certain distance from the channel bed. Ground vibration detectors are installed outside of the flow path, usually along the banks of the torrent or on the surrounding valley slopes, in order to avoid damage or even complete destruction. Seismic networks, however, are also prone to detect other earth surface processes that can be confused with the passage of a debris flow. Recognizing these other processes is important, particularly when the seismic network is used for warning purposes and not only for monitoring. To this aim, two seismic networks were installed in two instrumented basins located in the Italian Alps: the Marderello (Western Italian Alps) and the Gadria (Eastern Italian Alps). Both networks were designed for debris flow monitoring purposes and for testing warning algorithms. The seismic recordings of torrential processes that occurred at different distance from the monitoring networks, within and outside the monitored channels, are presented and discussed. It was found that knowledge of the waveform that these different processes produce is critical to the successful design and implementation of seismic networks for debris flow warning. The output of the seismic devices commonly employed for the monitoring of debris flows, such as geophones and seismometers, is a voltage that is directly proportional to the ground vibration velocity. The output signal in analogical form is usually digitalized at a fixed sampling frequency to be opportunely processed. The processing is performed to both reduce the amount of data to be stored in a data-logger and to reveal the main features of the phenomenon that are not immediately detectable in the raw signal, such as its main front, eventual subsequent surges, the wave form and so on. The processing also allows a better and sounder development of algorithms, when seismic devices are employed for warning purposes. However, the processing of the raw signal alters in different ways the original raw data, depending on the processing method adopted. This may consequently limit or reduce the efficacy of the warning. To this aim, the methods of amplitude and impulses have been also applied to some seismic recordings obtained in the instrumented basin of Illgraben (Switzerland) and Chalk Cliffs (USA). Data from two years of seismic monitoring performed in the Gadria basin are presented and discussed, together with the warning algorithm integrated in the system since 2014. In summer 2014 the alarm was correctly triggered by a debris flow and a red light was activated 3 minutes before the passage of the main front through the cross-section where this experimental semaphore is installed. The alarm lasted for the whole duration of the flow, correctly switching off after 20 minutes. Testing the algorithm on the whole seismic dataset recorded during both summers 2013 and 2014, the two debris flows event occurred in this time interval were correctly detected and only 3 false alarm where produced. After numerical simulation, it was possible to state that to eliminate false alarms a non-simultaneousness criterion in the threshold triggering can be adopted. The analysis of the frequency content of the seismic traces recorded at Gadria also allowed to draw some preliminary conclusions on the spectral characterization of the debris flow phenomena through geophone data. Some future developments are also outlined: the next step in the definition of a more advanced and robust warning system would be possible integrating the spectral information in the warning algorithm.