Search

Rapid mass movements are a major threat in populated landscapes, as they can cause significant loss of life and damage civil infrastructure. Previous work has shown that using environmental seismology methods to monitor such mass movements and establish monitoring systems offers advantages over existing approaches. The first important step in developing an early warning system for rapid mass movements based on seismic signals is automatically detecting events of interest. Though the approach, such as short-term average to long-term average ratio (STA/LTA) and machine learning model, was introduced to detect events (e.g., debris flow and rockfall), it is still challenging to calibrate input parameters and migrate existing methods to other catchments. Detection of debris flows, for instance, is similar to anomaly detection if we consider the seismic stations recording background signals as an overwhelming majority condition. Benford's law describes the probability distribution of the first non-zero digits in numerical datasets, which provides a functional, computationally cheap approach to anomaly detection, such as fraud detection in financial data or earthquake detection in seismic signals. In this study, seismic signals generated by rapid mass movements were collected to check the agreement of the distribution of the first digit with Benford's law. Subsequently, we develop a computationally efficient and non-site-specific model to detect events based on Benford's law using debris flows from the Illgraben, a Swiss torrent, as an example. Our results show that seismic signals generated by high-energy mass movements, such as debris flows, landslides, and lahars, follow Benford's law, while those generated by rockfall and background signals do not. Furthermore, our detector performance in picking debris-flow events is comparable to a published random forest and seismic network-based approach. Our method can be applied at other sites to detect debris-flow events without additional calibration and offers the potential for real-time warnings.

In order to reduce the societal impact of mass-wasting events, we need observations to investigate the factors that control slope failure, such as the state of crack propagation along a failure plane. However, usually the failure plane is not accessible in-situ. Hence, cracks have to be monitored indirectly, for example using seismic methods.We analysed the data from a seismometer array in the Illgraben catchment, Switzerland, that had registered a series of crack propagation and mass-wasting events, leading to a main event that happened on 2 January 2013. We used a state-of-the-art machine learning technique based on hidden Markov models to detect and classify the seismic signals of crack events. We obtained the temporal evolution of three signal types: (1) single crack signal, (2) rock avalanche and (3) rockfall activity due to debris remobilization. The temporal evolution of the number of cracks showed a linear trend in the weeks prior to the main mass-wasting event and, in the hours preceding the main event, a sigmoidal exponential growth. Using these observations, we propose a mechanistic model to describe the rupture of the failure plane. The model considers the internal parameter of the total crack boundary length as the primary control on failure plane evolution, in addition to the previously suggested crack propagation velocity control parameter. According to this model, internal parameters appear to be the dominant control for the failure plane growth at a slope scale.

International audience; This work focuses on the use of thin-layer models for simulating fast gravitational flows for hazard assessment. Such simulations are sometimes difficult to carry out because of the uncertainty on initial conditions and on simulation parameters. In this study, we aggregate various field data to constrain realistic initial conditions and to calibrate the model parameters. By using the SHALTOP numerical code, we choose a simple and empirical rheology to model the flow (no more than two parameters), but we model more finely the geometrical interactions between the flow and the topography. We can thus model both a rock avalanche, and the subsequent remobilization of the deposits as a high discharge debris flow.Using the Prêcheur river catchment (Martinique, Lesser Antilles) as a case study, we focus on extreme events with a high potential to impact populations and infrastructures. We use geological and geomorphological data, topographic surveys, seismic recordings and granulometric analysis to define realistic simulation scenarios and determine the main characteristics of documented events. The latter are then reproduced to calibrate rheological parameters. With a single rheological parameter and the Coulomb rheology, we thus model the emplacement and main dynamic characteristics of a recent rock avalanche, as well as the travel duration and flooded area of a documented high discharge debris flow. Then, in a forward prediction simulation, we model a possible 1.9x10^6 m^3 rock avalanche, and the instantaneous remobilization of the resulting deposits as a high-discharge debris flow. We show that successive collapses allow to better reproduce the dynamics of the rock avalanche, but do not change the geometry of the final deposits, and thus do not influence the initial conditions of the subsequent debris flow simulation. A progressive remobilization of the materials slows down the debris flow and limits overflow, in comparison to instantaneous release. However, we show that high discharge debris flows, such as the one considered for model calibration, are better reproduced with an instantaneous initiation. The range of travel times measured for other significant debris flows in the Prêcheur river is consistent with our simulation results, with various rheological parameters and the Coulomb or Voellmy rheology.

Bedload transport is a key process in fluvial morphodynamics, but difficult to measure. The advent of seismic monitoring techniques has provided an alternative to in‐stream monitoring, which is often costly and cannot be utilized during large floods. Seismic monitoring is a method requiring several steps to convert seismic data into bedload flux data. State‐of‐the‐art conversion approaches exploit physical models predicting the seismic signal generated by bedload transport. However, due to a lack of well‐constrained validation data, the accuracy of the resulting inversions is unknown. We use field experiments to constrain a seismic bedload model and compare the results to high‐quality independent bedload measurements. Constraining the Green's function (i.e., seismic ground properties) with an active seismic survey resulted in an average absolute difference between modeled and empirically measured seismic bedload power of 11 dB in the relevant frequency band. Using generically estimated Green's function parameters resulted in a difference of 20 dB, thus highlighting the importance of using actual field parameters. Water turbulence and grain hiding are unlikely to be the cause of differences between field observations and our analysis. Rather, they may be either due to the inverted model being particularly sensitive to the coarse tail of the grain‐size distribution, which is least well constrained from field observations, or due to the seismic model underestimating effects of the largest grains., Key Points: Constrained seismic model inversion reveals dominant control of largest grain size on bedload flux estimates Measuring ground properties from a seismic experiment allows improving spectrogram fit by an order of magnitude in seismic power Considering all field‐constrained parameters results in bedload flux overestimation by two orders of magnitude, Israel Science Foundation (ISF) http://dx.doi.org/10.13039/501100003977, NSF‐BSF

This work focuses on the use of thin-layer models for simulating fast gravitational flows for hazard assessment. Such simulations are sometimes difficult to carry out because of the uncertainty on initial conditions and on simulation parameters. In this study, we aggregate various field data to constrain realistic initial conditions and to calibrate the model parameters. By using the SHALTOP numerical code, we choose a simple and empirical rheology to model the flow (no more than two parameters), but we model more finely the geometrical interactions between the flow and the topography. We can thus model both a rock avalanche, and the subsequent remobilization of the deposits as a high discharge debris flow.Using the Prêcheur river catchment (Martinique, Lesser Antilles) as a case study, we focus on extreme events with a high potential to impact populations and infrastructures. We use geological and geomorphological data, topographic surveys, seismic recordings and granulometric analysis to define realistic simulation scenarios and determine the main characteristics of documented events. The latter are then reproduced to calibrate rheological parameters. With a single rheological parameter and the Coulomb rheology, we thus model the emplacement and main dynamic characteristics of a recent rock avalanche, as well as the travel duration and flooded area of a documented high discharge debris flow. Then, in a forward prediction simulation, we model a possible 1.9x106 m3 rock avalanche, and the instantaneous remobilization of the resulting deposits as a high-discharge debris flow. We show that successive collapses allow to better reproduce the dynamics of the rock avalanche, but do not change the geometry of the final deposits, and thus do not influence the initial conditions of the subsequent debris flow simulation. A progressive remobilization of the materials slows down the debris flow and limits overflow, in comparison to instantaneous release. However, we show that high discharge debris flows, such as the one considered for model calibration, are better reproduced with an instantaneous initiation. The range of travel times measured for other significant debris flows in the Pr\^echeur river is consistent with our simulation results, with various rheological parameters and the Coulomb or Voellmy rheology.

Constraining bedload flux in rivers is a challenging objective, especially when the data need to be continuous, beyond-point estimates. Seismometers are potentially valuable alternatives to in‐stream devices, which involve extensive measurement infrastructure or labour‐intensive manual sampling that can be potentially dangerous. We present a Monte Carlo-based inverse approach to deducing hydraulic and bedload transport dynamics continuously, with high temporal resolution, from seismic data, that averages the system’s behaviour over tens of metres. Water depths and bedload fluxes can be reproduced with average deviations of 0.10 m and 0.02 kg/sm, respectively. The method is validated against synthetic data sets and independently measured metrics from several challenging streams: we show applications of the technique from a flash flood-dominated catchment in Israel (Nahal Eshtemoa), from an ice-covered subarctic river (Sävarån, Sweden), and from a typhoon-driven major mountain river in Taiwan (Liwu River). The presented approach is a generic method implemented in the R package ‘eseis’ that can be used with off-the-shelf seismic equipment, installed at safe distances from potentially hostile conditions with minimum site disturbance.

Quantification of bedload flux along with its boundary conditions is essential to advance our understanding of rivers and to reduce human and economic threats. However, gaining continuous high resolution empirical data is challenging. Seismic sensors can provide time-resolved quantitative bedload flux data, given adequate data processing. The mechanistic model by Tsai et al. (2012) predicts the power spectral density (PSD) of Rayleigh waves caused by impacts of saltating particles on the river bed, allowing to invert seismic signals to obtain bedload flux. Here we test the robustness of bedload flux inversions of seismic observations against in-situ continuous monitoring of bedload flux and select bedload grain sizes made in the Nahal Eshtemoa, Israel. Proper testing is further ensured by wave propagation (the Green’s function) being fully constrained from active seismic survey experiments. We find that there is a discrepancy of approximately one order of magnitude between the measured and reconstructed bedload flux. We support that this discrepancy can be due to the largest grains, though constituting an infinitely small fraction, generating considerable seismic signals but not being caught by the bedload samplers. It is also possible that this discrepancy is due to model simplification regarding grains in motion. Based on these findings we support that seismic observations may be complementary rather than redundant to in-situ measured bedload flux, because they may give constraints on the fraction of large grains, which is challenging to monitor otherwise.

Rivers are the fluvial conveyor belts routing sediment across the landscape. While there are proper techniques for continuous estimates of the flux of suspended solids, constraining bedload flux is much more challenging, typically involving extensive measurement infrastructure or labour-intensive manual measurements. Seismometers are potentially valuable alternatives to in-stream devices, delivering continuous data with high temporal resolution on the average behaviour of a reach. Two models exist to predict the seismic spectra generated by river turbulence and bedload flux. However, these models require estimating a large number of parameters and the spectra usually overlap significantly, which hinders straightforward inversion. We provide three functions contained in the R-package eseis that allow generic modelling of hydraulic and bedload transport dynamics from seismic data using these models. The underlying Monte Carlo approach creates lookup tables of potential spectra, which are compared against the empirical spectra to identify the best fitting solutions. The method is validated against synthetic data sets and independently measured metrics from the Nahal Eshtemoa, Israel, a flash flood dominated ephemeral gravel bed river. Our approach reproduces the synthetic time series with average absolute deviations of 0.01--0.04 m (water depth, ranging between 0--1 m) and 0.00--0.04 kg/sm (bedload flux, ranging between 0--4 kg/sm). The example flash flood water depths and bedload fluxes are reproduced with respective average deviations of 0.10 m and 0.02 kg/sm. Our approach thus provides generic, testable and reproducible routines for a quantitative description of key metrics, hard to collect by other techniques in a continuous and representative manner.

Continuous positive airway pressure (CPAP) is a useful treatment for patients with acute cardiogenic pulmonary oedema (CPE). However, its usefulness in the out-of-hospital setting has been poorly investigated and only by small and single-centre studies. We designed a multicentre randomised study to assess the benefit of CPAP initiated out of hospital. A total of 207 patients with CPE were randomly allocated by emergency mobile medical units to receive either standard treatment alone or standard treatment plus CPAP. CPAP was maintained after admission to the intensive care unit (ICU). Inclusion criteria were orthopnoea, respiratory rate greater than 25 breaths/min, pulse oximetry less than 90% in room air and diffuse crackles. The primary end point was assessed during the first 48 h and combined: death, presence of intubation criteria, persistence of either all inclusion criteria or circulatory failure at the second hour or their reappearance before 48 h. Absence of all criteria defined successful treatment. CPAP was used for 60 min [40, 65] (median [Q1, Q3]) in the pre-hospital setting and 120 min [60, 242] in ICU and was well tolerated in all patients. Treatment was successful in 79% of patients in the CPAP group and 63% in the control group (p = 0.01), especially for persistence of inclusion criteria after 2 h (12 vs. 26%) and for intubation criteria (4 vs. 14%). CPAP was beneficial irrespective of the initial PaCO2 or left ventricular ejection fraction. Immediate use of CPAP in out-of-hospital treatment of CPE and until CPE resolves after admission significantly improves early outcome compared with medical treatment alone.