Logic Tree Research Articles

Trimming the UCERF3-TD logic tree: Model order reduction for an earthquake rupture forecast considering loss exceedance

The Uniform California Earthquake Rupture Forecast version 3-Time Dependent depicts California’s seismic faults and their activity. Its logic tree has 5760 leaves. Considering 30 more model combinations related to ground motion produces 172,800 distinct models representing so-called epistemic uncertainties. To calculate risk to a portfolio of buildings, one also considers millions of earthquakes and spatially correlated ground-motion variability. We offer a tree-trimming technique that retains the probability distribution of portfolio loss and identifies the leading sources of uncertainty for further study. We applied it to a California statewide building portfolio and various levels of nonexceedance probability between one in 100 and one in 2500. We trimmed the logic tree from 172,800 leaves to as few as 15. The result: a supercomputer that would otherwise run 24 h to estimate the distribution of one-in-250-year loss can calculate it in moments with the reduced-order model. Others can use the reduced-order model to calculate risk to different California portfolios, and scientists can prioritize study to reduce the remaining epistemic uncertainty.

Perawatan Preventive Maintenance Mobil Satlantas Model Sedan Mazda dengan Metode Reliability Centered Maintenance pada Polres Kota X

The aim of this research is to find out the preventive maintenance of Mazda Sedan Model Traffic Cars using the reliability centered maintenance method at the City functional block diagram, determining system function and functional failure, failure mode and effect analysis, logic tree analysis, and task selection. The results of this research are that the maintenance system for traffic vehicle units carries out routine check-ups every 5,000 kilometers or every 3 months and for brake and tire components every 10,000 kilometers or 6 months, the damage that appears with the current maintenance system is the break caliper (right front), hose (water pump), and ball joint, factors that arise with current maintenance problems are safety problems, proposed improvements that need to be made to resolve the problem are preventive maintenance for the break caliper (right front), hose (water pump), and ball joint every 11000 hours or 458 days.

Verifying chip designs at RTL level

As chip designs become increasingly complex, the potential for errors and defects in circuits inevitably rises, posing significant challenges to chip security and reliability. This study investigates the use of the SAT-based bounded model checking (BMC) for Propositional Projection Temporal Logic (PPTL) to verify Verilog chip designs at the register transfer level (RTL). To this end, we propose an algorithm to implement automated extraction of state transfer relations from AIGER netlist and construction of Kripke structure. Additionally, we employ PPTL with the full regular expressiveness to describe the circuit properties to be verified, especially the periodic repetitive properties. This is not possible with Linear Temporal Logic (LTL) and Computational Tree Logic (CTL). By combining the PPTL properties with finite system paths and transforming them into conjunctive normal forms (CNFs), we utilize an SAT solver for verification. Experimental results demonstrate that our verification tool, SAT-BMC4PPTL, achieves higher verification efficiency and comprehensiveness.

Assessment of Seismic Potential Impacts of an Mw 8.4 Hypothetical Earthquake in Central Nepal Province

The national capital of Nepal is situated on a lacustrine sediment basin. The country has had major seismic events that have resulted in significant damage to structures, human casualties, and substantial economic losses. Mitigating seismic risk is a challenging problem in Nepal due to poor construction practices, no enforcement of seismic safety guidelines, and a lack of awareness in the public. Seismic risk mitigation is essential in improving seismic resistance of buildings, and in reducing the economic loss and casualties in the forthcoming seismic events. The scientific results of earthquake loss estimation studies will lead to improve the policies towards seismic resilience.The current research uses the SELENA (Seismic Loss Estimation using a Logic Tree Approach) tool to explore the seismic damage to buildings, human loss, and seismic risk in the 11 districts due to a scenario earthquake. The seismic risk of the study region due to the scenario earthquake is determined through fragility functions. The expected economic losses vary from 0.1 to 0.6 million dollars, and the possible casualties range from 1,000 to 5,000. The outcome of the study will be helpful for the local authorities and policymakers to take mitigation measures for the existing buildings.

AnxietyNet: social media trends with fuzzy logic trees

In the digital age, social media platforms have become integral parts of daily life, facilitating communication, information sharing, and community building on a global scale. However, alongside the benefits of connectivity, concerns regarding public anxiety within these virtual communities have emerged. This research paper delves into the phenomenon of public anxiety on social media and proposes a novel approach utilizing fuzzy tree logic to analyze and address this issue. Through a comprehensive review of literature, theoretical frameworks, and practical applications, this paper aims to provide insights into the complex dynamics of public anxiety and offer potential solutions leveraging machine learning techniques. The proposed methodology involves the development of a cascading model to dynamically compute individual anxiety scores within social media networks. By analyzing structural influences and user interactions, this model aims to provide a nuanced understanding of anxiety dynamics within online communities. Additionally, a probabilistic model is designed to measure anxiety scores of social network messages, utilizing a generalized user profile. The integration of fuzzy tree logic enables the construction of a tree structure to effectively compute anxiety scores, offering a more flexible and human-like decision-making process. Preliminary experiments demonstrate the effectiveness of the proposed approach in capturing and analyzing public anxiety on social media platforms. The fuzzy tree model shows promise in accurately assessing anxiety levels and identifying key factors contributing to social anxiety within online communities. Moreover, the integration of machine learning techniques enhances the scalability and adaptability of the model, allowing for real-time analysis and proactive intervention strategies. This research underscores the importance of addressing public anxiety on social media through innovative analytical frameworks. By leveraging fuzzy tree logic and machine learning techniques, we can gain deeper insights into the dynamics of anxiety within online communities and develop targeted interventions to support individuals experiencing distress. Collaborative efforts in this endeavor hold the potential to foster societal well-being and resilience in the digital age.

Preventive Maintenance Planning for Machines Computer Numerical Control (CNC) Type Hurco Vmx24 to Improve Reliability Using the Reability Centered Maintenance (RCM) Method (Case Study)

Abstract: This paper presents the results of the analysis of Computer Numerical Control (CNC) machine components that result in decreased machine efficiency in the production system with the Reliability Centered Maintenance (RCM) approach. Utilized to improve the performance of machine components. The machine has not been maintained regularly so that at any time the machine can experience damage which results in losses to the company's income. so that an analysis is designed for maintenance planning management activities with the methodology of identifying and analyzing failure factors using the Failure Mode Effect Analysis (FMECA) worksheet, Logic Tree Analysis (LTA), and finally the Action Road Map. The results of the analysis carried out are divided into three categories of maintenance actions, namely, Condition Direction (CD) maintenance, Time Directed (TD) maintenance, Finding Failure (FF) maintenance with a periodic checking interval of 2 times within a period of 12 days.

Impact of seismotectonic source zoning and seismic parameter: Sensitivity study on probabilistic seismic hazard assessment

The demarcation of seismic area source zones plays a pivotal role in probabilistic seismic hazard assessment (PSHA). In this study, we developed five distinct seismotectonic area source models, integrating large and small areas based on geological, tectonic, and seismological data. Logic tree methodology is utilized in FRISK88M for PSHA. Analysis revealed higher hazards in models with larger area sources compared to those with smaller ones. Hazard deaggregation of seismotectonic model-5 (SM5) is examined and sensitivity analysis of various logic tree nodes is investigated in terms of percent deviation w.r.t mean hazard at various spectral and return periods. The findings underscore that recurrence methods fall into "insensitive level" (≤5%), source mechanism and a-value distribution belong to "low sensitive level" (>5–15%), magnitude distribution lies in "medium sensitive level" (>15–25%), b-value distribution belongs to "high sensitive level" (>25–35%), and ground motion prediction equations (GMPEs) and depth distribution falls into "very high sensitive level" (>35%) in relation to seismic hazard.

Forecasting In-Flight Icing over Greece: Insights from a Low-Pressure System Case Study

Forecasting in-flight icing conditions is crucial for aviation safety, particularly in regions with variable and complex meteorological configurations, such as Greece. Icing accretion onto the aircraft’s surfaces is influenced by the presence of supercooled water in subfreezing environments. This paper outlines a methodology of forecasting icing conditions, with the development of the Icing Potential Algorithm that takes into consideration the meteorological scenarios related to icing accretion, using state-of-the-art Numerical Weather Prediction model results, and forming a fuzzy logic tree based on different membership functions, applied for the first time over Greece. The synoptic situation of an organized low-pressure system passage, with occlusion, cold and warm fronts, over Greece that creates dynamically significant conditions for icing formation was investigated. The sensitivity of the algorithm was revealed upon the precipitation, cloud type and vertical velocity effects. It was shown that the greatest icing intensity is associated with single-layer ice and multi-layer clouds that are comprised of both ice and supercooled water, while convectivity and storm presence lead to also enhancing the icing formation. A qualitative evaluation of the results with satellite, radar and METAR observations was performed, indicating the general agreement of the method mainly with the ground-based observations.

Estimating the Maximum Magnitude of Induced Earthquakes in the Groningen Gas Field, the Netherlands

ABSTRACT Gas production in the Groningen field in the northern Netherlands began in 1963. Since 1991, many production-induced earthquakes have occurred—the largest reaching magnitude M 3.5 in 2012. In October 2023, the Dutch government closed the field—a decision that was justified by the potential threat of larger earthquakes occurring with continued gas extraction. This outcome highlights how the estimation of maximum magnitude for induced earthquakes can have important implications for risk perception by regulators and the public. Thus, the estimation of maximum magnitude for induced earthquakes that are not physically realizable can lead to unwarranted conservatism. Over almost three decades, considerable effort has been invested by several researchers to estimate the maximum magnitude of induced earthquakes in Groningen. These efforts culminated in international workshops focusing exclusively on this question in 2016 and 2022. The purpose of these workshops was to inform the decision-making of an expert panel charged with formulating a logic tree for the maximum magnitude of Groningen earthquakes. We argue that the evaluation of the final distribution of maximum magnitudes is overly conservative in terms of the largest events induced in the reservoir and, most importantly, the likelihood of triggering even larger tectonic earthquakes outside the gas-producing reservoir. Given the serious and far-reaching consequences of maximum magnitude estimates, we reconsider the approach of these assessments to draw lessons that may be critical for future energy-related projects associated with induced seismicity.

A Benchmarking Method to Rank the Performance of Physics-Based Earthquake Simulations

Abstract Physics-based earthquake simulators are an increasingly popular modeling tool in earthquake forecasting for seismic hazard as well as fault rupture behavior studies. Their popularity comes from their ability to overcome completeness limitations of real catalogs, and also because they allow reproducing complex fault rupture and interaction patterns via modeling the physical processes involved in earthquake nucleation and propagation. One important challenge of these models revolves around selecting the physical input parameters that will yield the better similarity to earthquake relationships observed in nature, for instance, the frictional parameters of the rate-and-state law—a and b—or the initial normal and shear stresses. Because of the scarcity of empirical data, such input parameters are often selected by trial–error exploration and predominantly manual model performance analyses, which can overall be time consuming. We present a new benchmarking approach to analyze and rank the relative performance of simultaneous earthquake simulation catalogs by quantitatively scoring their combined fit to three reference function types: (1) earthquake-scaling relationships, (2) the shape of the magnitude–frequency distributions, and (3) the rates of the surface ruptures from paleoseismology or paleoearthquake occurrences. The approach provides an effective and potentially more efficient approximation to easily identify the models and input parameter combinations that fit more closely to earthquake relations and behavior. The approach also facilitates the exhaustive analysis of many input parameter combinations, identifying systematic correlations between parameters and model outputs that can potentially improve the overall understanding of the physics-based models. Finally, we demonstrate how the method results agree with the published findings in other earthquake simulation evaluations, a fact that reinforces its overall usefulness. The model ranking outputs can be useful for subsequent analyses, particularly in seismic hazard applications, such as the selection of appropriate earthquake occurrence rate models and their weighting for a logic tree.

The 2023 Alaska National Seismic Hazard Model

US Geological Survey (USGS) National Seismic Hazard Models (NSHMs) are used extensively for seismic design regulations in the United States and earthquake scenario development, as well as risk assessment and mitigation for both buildings and infrastructure. This 2023 update of the long-term, time-independent Alaska NSHM includes substantial changes to both the earthquake rupture forecast (ERF) and ground motion models (GMMs). The ERF includes numerous additions to the finite-fault model, considers two deformation models, and introduces updated declustering and smoothing algorithms in the gridded background seismicity model. For the Alaska–Aleutian subduction zone, megathrust earthquakes occur on an updated structural and segmentation model, and the moment magnitude (M) 8+ rupture and rate model include a logic tree branch that considers slip rates derived from geodetic models of interface coupling. The megathrust model considers multiple models of down-dip width, and magnitudes are computed using newly developed scaling relations. For subduction intraslab events and subduction interface events with M < 7, the 2023 update uses a smoothed seismicity model with rupture depths derived from Slab2. The 2023 model updates GMMs in all tectonic settings using the recently published Next Generation Attenuation Subduction (NGA-Sub) GMMs for subduction interface and intraslab events, and the NGA-West2 GMMs for active crustal settings. Collectively, additions and updates to the Alaska NSHM result in hazard increases across most of south-central Alaska relative to the previous model, published in 2007. These changes are primarily due to the adoption of updated rate models for the large-magnitude interface events and the NGA-Sub GMMs that have much higher aleatory variability (sigma), consistent with global observations, and that include models of epistemic uncertainty.

Regional earthquake-induced landslide assessments for use in seismic risk analyses of distributed gas infrastructure systems

Earthquake-induced landslides are associated with significant risks to human lives and infrastructure. Spatially distributed infrastructure systems, such as pipelines, power lines, and transportation networks, are at particular risk to seismic landslides due to their large spatial extent. To conduct a comprehensive seismic landslide risk assessment for these systems, there is the need to evaluate the seismic landslide characteristics (i.e., location, size, displacement, direction) on a broad, regional scale. This paper presents a framework for seismic landslide analysis that provides this information for subsequent risk assessments. The approach computes seismic landslide displacements using a sliding block approach while accounting for the uncertainties in the input variables and displacement models using a logic tree. The computed displacements are then aggregated based on geomorphic landforms to define landslide zones. For each landslide zone, the statistical distributions of landslide features, such as landslide size, displacement level, and direction of movement, are defined. These attributes are presented in a format that can be integrated with fragility models for distributed infrastructure systems to quantify risk on a regional scale. The application of the approach is demonstrated through assessments for gas pipeline networks across the state of California in the United States.

Probabilistic Seismic Hazard Assessment of the Southwestern Region of Saudi Arabia

In relation to its rapid infrastructure expansion, exemplified by projects like the Najran Valley Dam or the rehabilitation of agricultural terraces, Saudi Arabia stands out among the Arabian Gulf nations. To mitigate the earthquake-related risks effectively, it is imperative to conduct an exhaustive analysis of its natural hazards. The southwesternmost region of Saudi Arabia is the main subject area of this study for the probabilistic seismic hazard assessment (PSHA), which aims to identify the peak ground acceleration (PGA) and spectral acceleration (SA) values. The investigation encompasses a 10% and 5% probability of occurrence over a 50-year exposure time for both B/C and C NEHRP soils. In order to take into account the earthquake activity that takes place in the vicinity of the Red Sea Rift, which in fact may have an impact on the seismic hazard in this active tectonic region, different seismic source zones were especially designed for this evaluation. Various characteristics such as the uncertainties related to the b-value, the expected maximum magnitude, and different ground motion prediction equations (GMPEs) were integrated using a logic tree scheme. Additionally, regression relationships between the computed ground motion values were established, and a novel design response spectrum was developed and recommended for several cities. Regarding the key findings, it is significant to highlight that the seismic hazard decreases towards the northeast, when moving away from the Red Sea Rift, confirming anticipated trends where proximity to the rift corresponds to increased seismic hazard. Notably, cities such as Farasan Island, Jazan, Al Qunfundhah, Al Lith and Al Birk present the highest observed hazard values among all the cities analyzed. For these cities, the obtained maximum SA values for both 475 and 975 years under B/C site conditions are as follows: 0.268 g and 0.412 g, 0.121 g and 0.167 g, 0.099 g and 0.150 g, 0.083 g and 0.135 g, and 0.066 g and 0.118 g, respectively. These results emphasize the crucial necessity of adequately evaluating and thoroughly updating the seismic hazard inherent to these particular areas to enhance the risk reduction and disaster readiness initiatives.

Probabilistic typhoon hazard and sensitivity analysis for nuclear power plant sites in Korea using logic tree

There has been an increase in typhoon intensity, and high-intensity typhoons can affect the safety of nuclear power plants (NPPs). Typhoons can damage the structures, systems, and components (SSCs) of NPPs, causing core damage and the release of radioactive materials into the environment. Therefore, design criteria for the effects of natural phenomena are demanded to ensure the safety-critical SSC functions of NPPs. Also, there is a need for safety assessments of typhoons in operational NPPs, which require typhoon hazard analysis. In this study, we derived typhoon hazards for NPPs in Korea using Logic Tree and Monte Carlo simulations. We performed sensitivity analysis on the typhoon hazards based on the distances (up to 100 km) from the NPP sites. Among other NPP sites, the typhoon hazard of the Kori site showed the highest wind speed, and the Hanul site showed the lowest wind speed. The uncertainty of the wind speed increased as the recurrence interval of the typhoon hazard increased. We compared the typhoon hazard in the areas of interest to that within 100 km. The range of the distance deviation for the Hanbit site's typhoon hazard was within μ±σ up to 75 km, while for the other sites, it was within μ±σ up to 100 km.

Prediction of diabetes mellitus using XGBoost model

With the number of health-threatening diseases and deaths on the rise, medical decision support systems continue to prove effective in improving the efficiency of physicians and other healthcare providers and supporting clinical decisions. Diabetes remains one of the leading diseases responsible for many deaths worldwide. Diabetes is characterized by elevated blood glucose levels, which can have serious consequences for other organs in the body. According to the International Diabetes Alliance (IDA), 382 million people currently have diabetes and this number is expected to double to 592 million by 2035. This paper proposes a clinical prediction model for diabetes based on machine learning (ML).As traditional ML models, we consider the most commonly used classifiers, the Logic Tree (LR) and XGBoost. We compare these ML models. On the other hand, we also use deep learning (DL) models and apply a fully convolutional neural network (DNN) for diabetes prediction and recognition. The proposed models were evaluated in the public Kaggle database: prediction performance at each level was analyzed using the Precision, Recall, AUC, accuracy and F1 metrics, and overall prediction efficiency was assessed using accuracy and its macro-average: DL, LR, The overall accuracy obtained by XGBoost was 91.88%, 88.75% and 97.12% respectively. The experimental results show that XGBoost is more effective at predicting diabetes than deep learning and LR methods.

Prediction of diabetes mellitus using XGBoost model

With the number of health-threatening diseases and deaths on the rise, medical decision support systems continue to prove effective in improving the efficiency of physicians and other healthcare providers and supporting clinical decisions. Diabetes remains one of the leading diseases responsible for many deaths worldwide. Diabetes is characterized by elevated blood glucose levels, which can have serious consequences for other organs in the body. According to the International Diabetes Alliance (IDA), 382 million people currently have diabetes and this number is expected to double to 592 million by 2035. This paper proposes a clinical prediction model for diabetes based on machine learning (ML).As traditional ML models, we consider the most commonly used classifiers, the Logic Tree (LR) and XGBoost. We compare these ML models. On the other hand, we also use deep learning (DL) models and apply a fully convolutional neural network (DNN) for diabetes prediction and recognition. The proposed models were evaluated in the public Kaggle database: prediction performance at each level was analyzed using the Precision, Recall, AUC, accuracy and F1 metrics, and overall prediction efficiency was assessed using accuracy and its macro-average: DL, LR, The overall accuracy obtained by XGBoost was 91.88%, 88.75% and 97.12% respectively. The experimental results show that XGBoost is more effective at predicting diabetes than deep learning and LR methods.

Ground-motion models for earthquakes occurring in the United Kingdom

This article presents models to predict median horizontal elastic response spectral accelerations for 5% damping from earthquakes with moment magnitudes ranging from 3.5 to 7.25 occurring in the United Kingdom. This model was derived using the hybrid stochastic-empirical method based on an existing ground-motion model for California and a stochastic model for the UK that was developed specifically for this purpose. The model is presented in two consistent formats, both for two distance metrics, with different target end-users. Firstly, we provide a complete logic tree with 162 branches, and associated weights, capturing epistemic uncertainties in the depth to the top of rupture, geometric spreading, anelastic path attenuation, site attenuation and stress drop, which is more likely to be used for research. The weights for these branches were derived using Bayesian updating of a priori weights from expert judgment. Secondly, we provide a backbone model with three and five branches corresponding to different percentiles, with corresponding weights, capturing the overall epistemic uncertainty, which is tailored for engineering applications. The derived models are compared with ground-motion observations, both instrumental and macroseismic, from the UK and surrounding region (northern France, Belgium, the Netherlands, western Germany and western Scandinavia). These comparisons show that the model is well-centred (low overall bias and no obvious trends with magnitude or distance) and that the branches capture the body and range of the technically defensible interpretations. In addition, comparisons with ground-motion models that have been previously used within seismic hazard assessments for the UK show that ground-motion predictions from the proposed model match those from previous models quite closely for most magnitudes and distances. The models are available as computer subroutines for ease of use.

Spatial Distribution of Earthquake Occurrence for the New Zealand National Seismic Hazard Model 2022

ABSTRACT We develop candidate hybrid models representing the spatial distribution of earthquake occurrence in New Zealand over the next 100 yr. These models are used within the onshore/near-shore, shallow component of the distributed seismicity model within the New Zealand National Seismic Hazard Model 2022. They combine a variety of spatially gridded covariates based on smoothed seismicity, strain rates, and proximity to mapped faults and plate boundaries in both multiplicative and additive hybrids. They were optimized against a standardized catalog of New Zealand earthquakes with magnitude M ≥ 4.95 and hypocentral depth ≤40 km from 1951 to 2020. We extract smoothed seismicity covariates using three different methods. The additive models are linear combinations of earthquake likelihood models derived from individual covariates. We choose three preferred hybrid models based on the information gain statistics, consideration of the ongoing Canterbury sequence and regions of low seismicity, and inclusion of the most informative covariates. Since the hazard model is designed for the next 100 yr, the preferred hybrid models are also combined with 20-year earthquake forecasts from the “Every Earthquake a Precursor According to Scale” model. Thus, in total, six hybrid spatial distribution candidates are advanced for sensitivity analyses and expert elicitation for inclusion in the final logic tree for the New Zealand National Seismic Hazard Model.

Pengendalian Kualitas Produk Gula Studi Kasus PT Madubaru PG. PS. Madukismo Yogyakarta

PG Madukismo is a sugar factory that was established in 1955 and is the only sugar factory in the province of the Special Region of Yogyakarta. This factory is located in Padokan Village, Tirtonirmolo Village, Kasihan District, Bantul Regency, Special Region of Yogyakarta. PT Madubaru, a company engaged in sugar cane processing, is a sugar and spirits factory located in the Yogyakarta area. Companies that produce cane sugar with SHS IA quality. Product defects are often found, resulting in a decrease in quality, therefore it is necessary to carry out quality control. This study uses the Failure Mode and Effect Analysis (FMEA) method to identify failure modes, Logic Tree Analysis (LTA) to determine the consequences of failure arising from failure modes and determine effective preventive maintenance policies for each machine. For the production quality control section, the Statistical Processing Control (SPC) method is used to analyze product defects to determine whether product defects are still within controllable limits or not. While the Failure Mode and Effect Analysis (FMEA) method is used to determine the failure mode in sugar production. For the quality control section on raw materials using the Economic Order Quantity (EOQ), Re-Order Point (ROP), Safety Stock (SS), and Total Inventory Cost (TIC) methods. The conclusion of this study is regarding the results of the analysis and calculations carried out with several settlement methods and there are several causes of production defects from 5 factors, such as: human factors, work methods, environment, raw materials and machines used. And provide solutions to companies in order to reduce defects by controlling product quality.

Modelling seismic ground motion and its uncertainty in different tectonic contexts: challenges and application to the 2020 European Seismic Hazard Model (ESHM20)

Abstract. Current practice in strong ground motion modelling for probabilistic seismic hazard analysis (PSHA) requires the identification and calibration of empirical models appropriate to the tectonic regimes within the region of application, along with quantification of both their aleatory and epistemic uncertainties. For the development of the 2020 European Seismic Hazard Model (ESHM20) a novel approach for ground motion characterisation was adopted based on the concept of a regionalised scaled-backbone model, wherein a single appropriate ground motion model (GMM) is identified for use in PSHA, to which adjustments or scaling factors are then applied to account for epistemic uncertainty in the underlying seismological properties of the region of interest. While the theory and development of the regionalised scaled-backbone GMM concept have been discussed in earlier publications, implementation in the final ESHM20 required further refinements to the shallow-seismicity GMM in three regions, which were undertaken considering new data and insights gained from the feedback provided by experts in several regions of Europe: France, Portugal and Iceland. Exploration of the geophysical characteristics of these regions and analysis of additional ground motion records prompted recalibrations of the GMM logic tree and/or modifications to the proposed regionalisation. These modifications illustrate how the ESHM20 GMM logic tree can still be refined and adapted to different regions based on new ground motion data and/or expert judgement, without diverging from the proposed regionalised scaled-backbone GMM framework. In addition to the regions of crustal seismicity, the scaled-backbone approach needed to be adapted to earthquakes occurring in Europe's subduction zones and to the Vrancea deep seismogenic source region. Using a novel fuzzy methodology to classify earthquakes according to different seismic regimes within the subduction system, we compare ground motion records from non-crustal earthquakes to existing subduction GMMs and identify a suitable-backbone GMM for application to subduction and deep seismic sources in Europe. The observed ground motion records from moderate- and small-magnitude earthquakes allow us to calibrate the anelastic attenuation of the backbone GMM specifically for the eastern Mediterranean region. Epistemic uncertainty is then calibrated based on the global variability in source and attenuation characteristics of subduction GMMs. With the ESHM20 now completed, we reflect on the lessons learned from implementing this new approach in regional-scale PSHA and highlight where we hope to see new developments and improvements to the characterisation of ground motion in future generations of the European Seismic Hazard Model.