Papers by Suzanne Lacasse
The prediction of landslide displacement is a key component of an early warning system to mitigat... more The prediction of landslide displacement is a key component of an early warning system to mitigate landslide risk. Most of the models for landslide displacement prediction are static models. Yet, landslides move dynamically. This paper proposes a novel dynamic model combining the wavelet transform and the multivariate long short-term memory neural network to predict landslide displacement. In the Three Gorges Reservoir Area (TGRA) in China, step-wise landslides have been observed. One such step-wise landslide, the Baijiabao landslide, was used as case study for this paper. The cumulated displacement was decomposed into a trend displacement, a periodic displacement and noise using the wavelet transform. The periodic displacement was predicted by the multivariate long short-term memory (LTSM) neural network considering various causal factors. For comparison, the static multivariate support vector machine (SVM) model and univariate LSTM model were also implemented. The results demonstrate that the multivariate LSTM model achieved higher prediction accuracy than the multivariate SVM and univariate LSTM models, and that the method is preferable for predicting the displacement of step-wise landslides in general, and for the TGRA in particular. RÉSUMÉ: La prédiction de l'étendue des glissements de terrain constitue un élément important pour la reduction du risque. La majorité des modèles existants pour glissements sont de nature statique. Un glissement se déplace cependant de manière dynamique. Cet article propose un nouveau modèle dynamique qui allie une transformation en ondelettes des signaux avec une analyse par réseau neuronal 'longue mémoire à court terme' (LSTM) avec multivariables. Autour du barrage "Three Gorges" en Chine, on a observé nombres de glissements "par étapes", dont le glissement de Baijiabao. Les déplacements cumulatifs sont composés d'une partie "trend" et une partie périodique. Le déplacement périodique a été modelé avec le réseau neuronal LSTM, considérant la précipitation et la variation du niveau du réservoir. Des comparaisons sont aussi faites avec un modèle statique, le modèle SVM (machine à vecteurs de support avec multi.variables) et un modèle univariable LSTM. Les résultats démontrent que le modèle LSTM avec multivariables fait un bien meilleur modelage que les deux autres modèles, et que cette méthode est préférable pour la prédiction des déplacements des glissements en général, et en particulier pour les glissements "par étapes" dans la région du barrage Three Gorges.
The paper describes two approaches for deriving the mean, standard deviation and probability dens... more The paper describes two approaches for deriving the mean, standard deviation and probability density function of the method uncertainty for an axial pile capacity calculation method. The focus of this paper is on estimating the statistical description of the method uncertainty parameters for a pile design method on the basis of performance of the method in predicting the capacities of high-quality pile load tests. The method uncertainty can have a strong influence on the safety level associated with the foundation design. Establishing the statistics of the "error" in a calculated capacity prediction (Qc) from the measured values of capacity (Qm) in pile load tests requires careful consideration of several factors. In particular, case studies demonstrated that only the pile load tests where the pile capacity method overpredicts the actual (measured) capacity are of interest. Therefore, with method uncertainty defined as Qm/Qc, the part of the cumulative distribution function where Qm/Qc < 1 should be fitted as well as possible. The possible dependence of the standard deviation of method uncertainty on pile penetration depth was also investigated in the derivation of method uncertainty statistics.
Natural Hazards, Jun 20, 2022
Upon the introduction of machine learning (ML) and its variants, in the form that we know today, ... more Upon the introduction of machine learning (ML) and its variants, in the form that we know today, to the landslide community, many studies have been carried out to explore the usefulness of ML in landslide research and to look at some classic landslide problems from an ML point of view. ML techniques, including deep learning methods, are becoming popular to model complex landslide problems and are starting to demonstrate promising predictive performance compared to conventional methods. Almost all the studies published in the literature in recent years belong to one of the following three broad categories: landslide detection and mapping, landslide spatial forecasting in the form of susceptibility mapping, and landslide temporal forecasting. In this paper, we present a brief overview of ML techniques, provide a general summary of the landslide studies conducted, in recent years, in the three above-mentioned categories, and make an attempt to critically evaluate the use of ML methods to model landslide processes. The paper also provides suggestions for future use of these powerful data-driven techniques in landslide studies.
Paper No. SOA-5 Statistics. reliability analyses and risk estimates can he very useful decision-m... more Paper No. SOA-5 Statistics. reliability analyses and risk estimates can he very useful decision-making tools in geotechnical problems. Yet the methods are little used in practice. The offshore and mining industry arc at the forefront for the usc of these approaches, having encouraged their use and sponsored research that has enabled the methods to he well-documented and of proven usefulness in the study of alternatives for design and decision-making in face of uncertainties. The paper presents a few case studies in diiTerent areas of geotechnical engineering and discusses the results that would have hccn nhtained without the use nf the risk approach. Special emphasis is given to dams and offshore structures, hoth piled and shallow foundations. The authors take a look at the reasons why the methods arc not used to a greater extent in praclice and make recommendations as to when and how one should uses such methods.
Engineering Geology
Tailings dams are commonly built incrementally to increase the storage capacity of the Tailings S... more Tailings dams are commonly built incrementally to increase the storage capacity of the Tailings Storage Facility (TSF), usually without interrupting the mining activities. Dam management practices, lack of knowledge on tailings behaviour and the poor performance of monitoring and management processes have resulted in disastrous tailings dam failures with human and economic losses, as well as huge environmental consequences to ecosystems and local communities. In the literature, correlation analyses have been carried out considering different variables: stored volume, released volume, runout distance, dam height, peak discharge. Several databases of tailings dam failure are available online, each with different levels of detail. This paper computes the statistics of tailings dam failures using an up-to-date database on failures and a catalogue of existing TSF. The existing correlations between stored and released volumes have been verified using a larger database. The new proposed regression analysis considers the functional relationship between released volume and characteristics of the dam such as height and stored volume (i.e., dam factor). The effect of construction type, fill material and failure mode on the released volume has also been evaluated as well as the frequency of tailings dam failure as function of the construction method. Tailing dams built using the upstream construction method turn out to be more prone to failure, and more susceptible to static and dynamic liquefaction. The new correlation provides more reliable estimates of the expected released volume as a function of dam height and stored volume and should prove useful for runout analyses and risk assessment of tailings dam failure. Finally, the analyses carried out show that there is no correlation between the water pond extension and the released volume.
Géotechnique, 2022
Rain-induced man-made slope failures pose great threats to public safety as most man-made slopes ... more Rain-induced man-made slope failures pose great threats to public safety as most man-made slopes are formed in densely populated areas. A critical step in managing landslide risks is to predict the time, locations and consequences of slope failures in future rainstorms. Based on comprehensive databases of in-service man-made slopes, rainstorms and landslides in Hong Kong during the past 35 years, a spatio-temporal landslide forecasting model for man-made slopes is developed in this study within a unified machine learning framework. With a storm-based data integration strategy and multiclass classification on landslide scales, the framework incorporates landslide time and consequences in landslide susceptibility mapping to successfully achieve spatio-temporal landslide forecasting. The machine learning-based landslide forecasting model is validated against historical landslide incidents both temporally and spatially and through a case study of the June 2008 storm; the model significa...
Understanding and Reducing Landslide Disaster Risk, 2020
The paper presents the result of an extensive study on the undrained holding capacity of suction ... more The paper presents the result of an extensive study on the undrained holding capacity of suction anchors with different length-to-diameter ratios. The clays used in this study have a linearly increasing undrained shear strength with depth. The paper proposes a ‘ready-to-use’ equation for the preliminary design of suction anchors in soft clays. To highlight the benefit of the design equation, three example projects where detailed design of suction anchors had already been carried out numerically are compared with the equation-based design. The comparisons show that the proposed equation can be used with confidence for the preliminary design of suction anchors in similar soil conditions. The paper also addresses the key assumptions and restrictions associated with the use of the design equation. RÉSUMÉ : L'article compare l'analyse numérique et analytique de la résistance d'ancrages à succion de différentes dimensions et installées dans de l'argile molle à grande profo...
Risk assessment and management of flow landslides require a reliable estimate of the runout of th... more Risk assessment and management of flow landslides require a reliable estimate of the runout of the landslide masses. This paper introduces empirical and analytical models for the prediction of the runout of flow landslides. The numerical model uses an extension of the Bing model in Eulerian coordinates with two-space dimensions and implements the full Herschel–Bulkley rheology to dynamically compute the depth of the moving material and shear layer. The models are validated by comparing them to the observed runout values for the Kattmarka flow landslide that took place in Norway in 2009. In particular, the analytical model, although still under development, shows promise.
Lecture Notes in Civil Engineering, 2019
Society and standards require more and more “risk-informed” decisions. The paper demonstrates the... more Society and standards require more and more “risk-informed” decisions. The paper demonstrates the potential of reducing risk by implementing reliability and risk concepts as a complement to conventional analyses. Reliability evaluations can range from qualitative estimates, simple statistical evaluations to full quantitative probabilistic modelling of the hazards and consequences. The paper first introduced recent innovative developments that help reduce risk. Risk assessment and risk management are briefly touched upon. An example of the application of the new stress testing method is given. The usefulness of the seminal (1969) Observational Method is discussed. The need for developing sustainable and holistic civil engineering solutions is also briefly mentioned. The paper concludes that reliability-based approaches provide useful complementary information, and enable the analysis of complex uncertainties in a systematic and more complete manner than deterministic analyses alone. There is today a cultural shift in the approach for design and risk reduction in our profession. Reliability and risk-based approaches will assist preparing sustainable engineering recommendations and making risk-informed decisions.
Society increasingly requires the engineer to quantify and manage the risk which people, property... more Society increasingly requires the engineer to quantify and manage the risk which people, property and the environment are exposed to. The role of the geotechnical engineering profession is to reduce exposure to threats, reduce risk and protect people. Hazard, reliability and risk approaches are excellent tools to assist the geotechnical engineer in design, selection of engineering foundation solutions and parameters and decision-making. The significance of factor of safety is discussed, and basic reliability and risk concepts are briefly introduced. The importance of designing with a uniform level of reliability rather than a constant safety factor prescribed in codes and guidelines is illustrated. The paper illustrates the use of the reliability and risk concepts with "real life" case studies, in particular for situations encountered for Nordic environments. The calculation examples are taken from a wide realm of geotechnical problems, including avalanche, railroad safety...
IOP Conference Series: Earth and Environmental Science, 2021
In 2015, a sudden landslide caused the failure of one of the pillars supporting the southern lane... more In 2015, a sudden landslide caused the failure of one of the pillars supporting the southern lanes of the Skjeggestad Bridge near Mofjellbekken on Expressway E18. The transportation corridor was closed to traffic for 17 months. To investigate the cause of the failure, an assessment of slope stability is necessary. Usually, limit equilibrium analysis of the middle two-dimensional (2D) cross-section is modelled. The Skjeggestad landslide geometry was not close to a 2D case, and three-dimensional (3D) modelling is more appropriate to analyse the slope. The paper calculates the stability of the slope that failed and compares the results of 3D finite element analyses with classical limit equilibrium and 2D finite element analyses. The analyses were run in the Novapoint GeoSuite Stability software. The soil parameters, including their statistical values, were obtained with the GeoSuite Soil Data Interpretation (SDI) module. A companion paper at this conference analyses the pillar neighbou...
The paper describes and validates the response surface approach developed to estimate the stresse... more The paper describes and validates the response surface approach developed to estimate the stresses and forces of a structure. The proposed approach eliminates the need for many large structural computations while accounting for environmental load combinations, and computes the stresses and forces in the members and joints with a simpler and more efficient relationship than conventional structural analysis.
Proceedings of the 7th International Symposium on Geotechnical Safety and Risk (ISGSR 2019), 2019
A new data fusion approach is presented for evaluating the vulnerability curves utilizing multi-s... more A new data fusion approach is presented for evaluating the vulnerability curves utilizing multi-source monitoring information. A Bayesian network with continuous variables for a shield driven tunnel involving the loading condition, structural parameters (e.g. stiffness in waist and top joint) and soil parameters (e.g. subgrade reaction and lateral earth pressure) was constructed. Field monitoring data are taken into the Bayesian network to update the uncertain variables. Fragility and vulnerability curves are then drawn using the updated parameters. An example of shield driven tunnel under extreme surcharging is presented to illustrate the proposed methodology. The method is able to integrate the field monitoring information based on the tunnel deformation mechanisms and update the soil and structural parameters.
Proceedings of the 6th International Symposium on Reliability Engineering and Risk Management, 2018
The paper describes two approaches for deriving the mean, standard deviation and probability dens... more The paper describes two approaches for deriving the mean, standard deviation and probability density function of the method uncertainty for an axial pile capacity calculation method. The focus of this paper is on estimating the statistical description of the method uncertainty parameters for a pile design method on the basis of performance of the method in predicting the capacities of high-quality pile load tests. The method uncertainty can have a strong influence on the safety level associated with the foundation design. Establishing the statistics of the "error" in a calculated capacity prediction (Qc) from the measured values of capacity (Qm) in pile load tests requires careful consideration of several factors. In particular, case studies demonstrated that only the pile load tests where the pile capacity method overpredicts the actual (measured) capacity are of interest. Therefore, with method uncertainty defined as Qm/Qc, the part of the cumulative distribution function where Qm/Qc < 1 should be fitted as well as possible. The possible dependence of the standard deviation of method uncertainty on pile penetration depth was also investigated in the derivation of method uncertainty statistics.
Proceedings of the 6th International Symposium on Reliability Engineering and Risk Management, 2018
Landslides are a major hazard in Hong Kong which can lead to loss of life, injury to people, or e... more Landslides are a major hazard in Hong Kong which can lead to loss of life, injury to people, or economic losses. This paper assesses the landslide risk in western part of Hong Kong Island under extreme rainstorms of 29%, 44%, 65% and 85% of the 24-h Probable Maximum Precipitation. The number of buildings affected and the total population inside the buildings have been identified. The vulnerability factor was evaluated as a function of travel angle and time of landslide occurrence. It was observed that the vulnerability factor increases as the travel angle increases for all the PMP levels. At 85% and 65% PMP, the occurrence of landslides and the risk they pose on buildings are much higher than those at 44% and 29% PMP. The lowest potential loss of life is observed in schools, hospitals, community centres and government buildings. The findings from this work suggest that the time of occurrence of landslides can significantly affect the distribution of potential loss of life in each building.
Uploads
Papers by Suzanne Lacasse