GEO Partner (Switzerland)

Rockfall runout and energy dissipation are controlled by complex interactions between block characteristics and terrain properties, yet simplified approaches remain widely used for rockfall hazard assessments in the praxis. This contribution revisits the rockfall energy line method (also known as the Pauschalgefälle method), presenting its conceptual basis and an online implementation tool for estimation of rockfall velocities and kinetic energies along slope profiles. The method projects an idealised energy line from the rockfall release area to the distal margin of the deposit zone, comparable to the shadow angle approach, with modified slope angles used to approximate terrain resistance due to surface roughness and vegetation.Despite its simplicity, the rockfall energy line method is commonly employed as a first-order estimate and plausibility check in rockfall hazard assessments. Here, we evaluate the method against data from controlled, real-world rockfall experiments and examine its performance relative to advanced numerical rockfall models. The comparison illustrates how this simplified energy-based approach can complement process-based simulations, particularly where field data are limited.Our benchmarking highlights some limitations of the current rockfall energy line method, particularly for large, idealised blocks travelling over smooth alpine meadow terrain. Based on these findings, we propose practical adaptations to the method that improve its applicability to extreme rockfall scenarios and provide guidance for its appropriate use in rockfall hazard assessment for the praxis.

Estimating extreme flood events with a return period of 1000 years or more is particularly challenging in ungauged catchments. Traditional methods often rely on statistical extrapolation of peak discharges or regionalized design values, both of which are subject to considerable uncertainty. Our study examines whether continuous hydrological simulations could offer a useful second opinion to these traditional methods. The continuous simulations are based on a model chain that starts with a stochastic weather generator, which produces long synthetic time series of precipitation and temperature. These time series then serve as input to a hydrological model. As there are no direct streamflow observations for ungauged catchments, the model parameters must be regionalized to realistically configure the hydrological model. Initial tests, in which gauged catchments were treated as ungauged, were promising, suggesting that continuous simulation using a stochastic weather generation combined with a hydrological model could provide a robust basis for estimating extreme floods in regions with limited data.

This chapter explores the theoretical foundations, differences, applications, and future directions of ensemble and hybrid machine learning techniques, with a particular emphasis on their application in healthcare. Ensemble learning combines multiple models to enhance accuracy and robustness by leveraging the strengths of individual models, using methods such as bagging, boosting, and stacking. Hybrid machine learning integrates diverse algorithms and paradigms to exploit their complementary advantages, leading to more powerful and adaptable models. We discuss the conceptual and methodological differences between these techniques, highlighting their unique strengths and limitations. Practical applications are examined across various domains, with a specific focus on healthcare applications such as disease diagnosis, medical image analysis, personalized medicine, and treatment optimization. A dedicated chapter on healthcare integration addresses how these advanced techniques can be tailored to meet specific challenges in the healthcare domain, enhancing the accuracy, efficiency, and effectiveness of medical practices. Key challenges such as computational complexity, model interpretability, and scalability are addressed, along with emerging research opportunities and technological advancements, including quantum computing and federated learning. The insights gained provide a comprehensive understanding of these advanced machine learning techniques, guiding their effective application and future development to tackle complex real-world problems, especially in healthcare.

Quantification of liver fat on computed tomography (CT) is often confounded by hepatic iron deposition and the use of iodinated contrast agents. This phantom study aimed to evaluate the feasibility and accuracy of quantifying liver fat content (LFC) in the presence of iron using spectral localizer radiographs acquired with photon-counting detector CT (PCD-CT). Sixteen liver phantoms were constructed using mixtures of liver tissue, fat, and iron to simulate 4 levels of LFC (0%, 10%, 30%, and 50%) and 4 levels of liver iron concentration (LIC: 0, 1.5, 3, and 6mg/mL). Five additional reference phantoms (containing fat only, water only, or water-iron solutions) were included. All phantoms were scanned on a clinical PCD-CT system using 3 tube current settings (10, 50, 300mA) to acquire spectral localizer radiography data. Material decomposition of high- and low-energy bin data yielded water and hydroxyapatite (HA) maps. HA values were analyzed as a function of LFC and LIC, and water values were correlated with corresponding HA values. Increasing LFC resulted in a linear decrease in HA values, consistent across all LIC levels (slopes=-0.0016 to -0.0023; mean=-0.0019; r=0.997 to 1.0). Conversely, increasing LIC caused a linear increase in HA values, independent of LFC (slopes=0.0147 to 0.017; mean=0.0156; r=0.978 to 1.0). When combined with water values in a 2-dimensional material space, these stable linear relationships enabled estimation of LFC irrespective of LIC. Findings were reproducible across all tube current settings. Spectral localizer radiographs from PCD-CT allow quantification of liver fat content even in the presence of iron deposition. If validated in vivo, this technique may enable low-threshold opportunistic screening for hepatic steatosis and iron overload from precontrast localizer scans.

When the public is asked to name the most globally sustainable industries, they typically respond with solar, wind, geothermal, or EVs. Yet, who would ever imagine the pulp & paper industry, which is often caricatured as a relic, is in fact one of the largest and most secretly sustainable manufacturing sectors. Its entire infrastructure and operations are built on replenishable forests and powered by renewable energy streams to produce recyclable, biodegradable products. The pulp & paper industry has a story that deserves retelling in the age of sustainability metrics and ESG frameworks. Our editorial embarks on a short simple journey to reframe pulp & paper not as a legacy industry, but as a model for sustainable manufacturing by using a clear, quantifiable system to demonstrate its global environmental impact.

We present a data-driven motion correction method for brain positron emission tomography (PET) using Monte Carlo simulations of a Derenzo phantom and measure FDG patient data. Our approach performs motion detection from listmode allowing us to remove motion while minimising the number of frames. Motion estimation is performed in image space while reconstruction of PET images is done using STIR library. The line profile results demonstrate recovery of the resolution for the motion-corrected images even for the 2 mm diameter spheres. Furthermore, the motion corrected patient images recovers more defined brain structures that originally suffered from motion.

<p>The presented footbridge is part of a vast urban project to redevelop the train station and the surrounding public spaces in Renens, a small community situated in the greater Lausanne area in Switzerland. It is the result of a design competition won in 2007 by the “Rayon vert” (Green Ribbon) consortium, a close collaboration between architects and engineers.</p><p>Throughout the project, the engineers were given significant creative freedom, which pushed the structural concept to the limit. This in turn strengthened the architectural concept, and the resulting design is both coherent and poetic. Discussing the engineering practice and vision with the architects is above all a way of bringing architecture and structure into dialogue, so as not to reduce the latter to its most basic load-bearing function, but rather to integrate it holistically into the architecture.</p>

<strong class="journal-contentHeaderColor">Abstract.</strong> Accurate forecasting of the risk posed by catastrophic failure of rock slopes requires estimates of the potential impact area and emplacement velocity. While most previous work in this context has focused on rock avalanche behaviour, recent and well documented case histories are showing that a more diverse range of landslide classes can occur. In the present work, we analyse two rock slope failures that occurred at Brienz/Brinzauls in Switzerland. These events initiated within 500 m of each other on the same slope, but emplaced with velocities that differed by five orders of magnitude. We describe the derivation and implementation of a GPU accelerated numerical model that can simulate emplacement velocities on the order of m/day. We then perform forensic back-analysis of the two case histories. Our results highlight the role of path material in controlling emplacement behaviour, as well as the effect of moderate changes in source material lithology. We argue that these cases can form the foundation of more accurate hazard and risk analyses at similar sites, where a wider range of potential future behaviour than is typical should be considered.

Abstract. Accurate forecasting of the risk posed by catastrophic failure of rock slopes requires estimates of the potential impact area and emplacement velocity. While most previous work in this context has focused on rock avalanche behaviour, recent and well documented case histories are showing that a more diverse range of landslide classes can occur. In the present work, we analyse two rock slope failures that occurred at Brienz/Brinzauls in Switzerland. These events initiated within 500 m of each other on the same slope, but emplaced with velocities that differed by five orders of magnitude. We describe the derivation and implementation of a GPU accelerated numerical model that can simulate emplacement velocities on the order of m/day. We then perform forensic back-analysis of the two case histories. Our results highlight the role of path material in controlling emplacement behaviour, as well as the effect of moderate changes in source material lithology. We argue that these cases can form the foundation of more accurate hazard and risk analyses at similar sites, where a wider range of potential future behaviour than is typical should be considered.