The global building exposure model is a mosaic of local and regional models with information regarding the residential, commercial, and industrial building stock at the smallest available administrative division of each country and includes details about the number of buildings, number of occupants, vulnerability characteristics, average built-up area, and average replacement cost. We aimed for a bottom-up approach at the global scale, using national statistics, socio-economic data, and local datasets. This model allows the identification of the most common types of construction worldwide, regions with large fractions of informal construction, and areas prone to earthquakes with a high concentration of population and building stock. The mosaic of exposure models presented herein can be used for the assessment of probabilistic seismic risk and earthquake scenarios. Information at the global, regional, and national levels is available through a public repository (https://github.com/gem/global_exposure_model), which will be used to maintain, update and improve the models.

The Development of a Global Seismic Risk Model was a mammoth undertaking that involved hundreds of people and for the first time presented a detailed view of seismic risk at the global scale. For some developing countries, this was the first time that a seismic risk map was produced, and the associated country profiles are being used by the local authorities.

The well-established Bath’s law states that the average magnitude difference between a mainshock and its strongest aftershock is roughly 1.2, independently of the size of the mainshock. The main challenge in calculating this value is the bias introduced by missing data points when the strongest aftershock is below the observed cut off magnitude. Ignoring missing values leads to a systematic error, because the data points removed are those with particularly large magnitude differences ∆M. The error is minimized, if we restrict the statistics to mainshocks at least two magnitude units above the cut-off, but then the sample size is strongly reduced. This work provides an innovative approach for modelling ∆M by adapting methods for time-to-event data, which often suffers from incomplete observation (censoring). In doing so, we adequately account for unobserved values and estimate a fully parametric distribution of the magnitude differences ∆M for M ą 6 mainshocks. Results show that magnitude differences are best modeled by the Gompertz distribution, and that larger ∆M are expected at increasing depths and higher heat flows. A simulation experiment suggests that ∆M is mainly driven by the number and the magnitude distribution of aftershocks. Therefore, in a second study, we modelled the variation of aftershock productivity in a stochastically declustered local catalog for New Zealand, using a generalized additive model approach. Results confirm that aftershock counts can be better modelled by a Negative Binomial than a Poisson distribution. Interestingly, there is indication that triggered earthquakes trigger themselves two to three times more aftershocks than comparable

This work investigates a novel approach combining numerical modelling and machine learning, aimed at developing an efficient procedure that can be used for large scale tsunami hazard and risk studies. Probabilistic tsunami hazard and risk assessment are vital tools to understand the risk of tsunami and mitigate its impact, guiding the risk reduction and transfer activities. Such large-scale probabilistic tsunami hazard and risk assessment require many numerically intensive simulations of the possible tsunami events, involving the tsunami phases of generation, wave propagation and inundation on the coast, which are not always feasible without large computational resources like HPCs. In order to undertake such regional PTHA for a larger proportion of the coast, we need to develop concepts and algorithms for reducing the number of events simulated and more rapidly approximate the simulation results needed. This case study for a coastal region of Japan utilizes a limited number of tsunami simulations from submarine earthquakes along the subduction interface to generate a wave propagation database at different depths, and fits these simulation results to a machine learning model to predict the water depth or velocity of the tsunami wave at the coast. Such a hybrid ML-physical model can be further coupled with an inundation scheme to compute the probabilistic tsunami hazard and risk for the onshore region.

We performed benefit-cost analysis to identify optimum retrofitting interventions for the two most vulnerable building typologies in Central America, unreinforced masonry and adobe, considering the direct costs due to building damage and the indirect costs associated with the injured and fatalities. We reviewed worldwide retrofitting techniques, selected those that could be applied in the region for these building types, and derived vulnerability functions considering the impact of each retrofitting intervention in the strength, stiffness, and ductility of the structures. Probabilistic seismic risk analyses were performed considering the original configuration of each building class, as well as the retrofitted version. We calculated average annual losses to estimate the annual savings due to the different structural interventions, and benefit cost ratios were estimated based on the associated cost of each retrofitting technique. Based on the benefit-cost analyses, for a 50-year time horizon and a 4% discount rate, retrofitting these building classes could be economically viable along the western coast of Central America.

Vitor Silva reflects on the current position of seismic risk assessment compared to its hazard counterpart, and posits that this discipline is expected to become common practice in disaster risk management, providing decision makers with valuable information not just about the current threat, but also how the impact of future disasters is expected to evolve. The growth of seismic risk assessment into its adult years will allow a more efficient design and implementation of risk mitigation measures. ultimately contributing to its main and only goal: the reduction of the human and economic losses caused by earthquakes.

This study proposes a framework to forecast the spatial distribution of population and residential buildings for the assessment of future disaster risk. The approach accounts for the number, location, and characteristics of future assets considering sources of aleatory and epistemic uncertainty in several time-dependent variables. The value of the methodology is demonstrated at the urban scale using an earthquake scenario for the Great Metropolitan Area of Costa Rica. Hundreds of trajectories representing future urban growth were generated using geographically weighted regression and multiple-agent systems. These were converted into exposure models featuring the spatial correlation of urban expansion and the densification of the built environment. The forecasted earthquake losses indicate a mean increase in the absolute human and economic losses by 2030. However, the trajectory of relative risk is reducing, suggesting that the long-term enforcement of seismic regulations and urban planning are effectively lowering seismic risk in the case of Costa Rica.

The development of building inventory is a fundamental step for the evaluation of the seismic risk at territorial scale. Census data are usually employed for building inventory in large scale application and their use requires suitable rules to assign buildings typologies to vulnerability classes, that is an exposure model specifc for the considered vulnerability model. Several exposure models are developed proposing class assignment rules that are calibrated on building typological data available from post-earthquake survey data. However, this approach has the drawback of being based on data from specifc geographic areas that have been hit by damaging earthquakes. Indeed, the distribution of building typologies can vary greatly for diferent areas of a country and the difusion of one building’s typology rather than another one may depend on the availability of construction material in the area, the evolution of construction techniques and the codes in force at the time of construction. This paper aims to improve the exposure modelling at regional scale, investigating the variability of masonry building typologies distribution. It proposes a methodology to recalibrate the exposure models at regional scale and evaluates the infuence of the improved characterization of regional vulnerability on damage and risk assessment. The study shows that the analysis of local building typologies may strongly impact on the evaluation of the seismic risk at territorial scale.

Major faults of the Tien Shan, Central Asia, have long repeat times, but fail in large (MwE 7+) earthquakes. In addition, there may be smaller, buried faults off the major faults which are not properly characterized or even recognized as active. These all pose hazard to cities along the mountain range front such as Almaty, Kazakhstan. Here, we explore the seismic hazard and risk for Almaty from specific earthquake scenarios. We run three historical-based earthquake scenarios (1887 Verny MwE 7.3, 1889 Chilik MwE 8.0 and 1911 Chon-Kemin MwE 8.0) on the current population and four hypothetical scenarios for near-field faulting. By making high-resolution Digital Elevation Models (DEMs) from SPOT and Pleiades stereo optical satellite imagery, we identify fault splays near and under Almaty. We assess the feasibility of using DEMs to estimate city building heights, aiming to better constrain future exposure datasets. Both Pleiades and SPOT-derived DEMs find accurate building heights of the majority of sampled buildings within error; Pleiades tri-stereo estimates 80% of 15 building heights within one sigma and has the smallest average percentage difference to field-measured heights (14%). A moderately sized MwE 6.5 earthquake rupture occurring on a blind thrust fault, under folding north of Almaty is the most damaging scenario explored here due to the modeled fault stretching under Almaty, with estimated 12,300E5,000 completely damaged buildings, 4,100 E 3,500 fatalities and an economic cost of 4,700 E 2,700 Million US dollars (one sigma uncertainty). This highlights the importance of characterizing location, extent, geometry, and activity of small faults beneath cities.

The incorporation of machine learning (ML) algorithms in earthquake engineering can improve existing methodologies and enable new frameworks to solve complex problems. In the present study, the use of artificial neural networks (ANNs) for the derivation of seismic vulnerability models for building portfolios is explored. Large sets of ground motion records (GMRs) and structural models representing the building stock in the Balkan region were used to train ANNs for the prediction of structural response, damage and economic loss conditioned on a vector of ground shaking intensity measures. The structural responses and loss ratios (LRs) generated using the neural networks were compared with results based on traditional regression models using scalar intensity measures in terms of efficiency, sufficiency, bias and variability. The results indicate a superior performance of the ANN models over traditional approaches, potentially allowing a greater reliability and accuracy in scenario and probabilistic seismic risk assessment.