The Urgent Computing Integrated Services for Earthquakes (UCIS4EQ) is proposed as a novel Urgent Computing (UC) seismic workflow that focuses on short-time reports of synthetic estimates of the consequences of moderate to large earthquakes. UC combines High-Performance Computing (HPC), High-Performance Data Analytics (HPDA), and optimized solvers to perform numerical simulations during or immediately after emergency situations, typically within a few minutes to a few hours. Complex edge-to-end UC workflows coordinate the execution of multiple model realizations to account for input and model uncertainties and can provide decision-makers with numerical estimates of the outcomes of emergency scenarios, such as earthquakes addressed by UCIS4EQ. UCIS4EQ is being driven toward operational maturity thanks to the technological and scientific developments within the eFlows4HPC project. Based on containerised micorservices, this workflow is fully orchestrated by the PyCOMPSs workflow manager to automatically prepare and manage physics-based deterministic simulation suites for rapid synthetic results. Through pre-computed and on-the-fly simulations, UCIS4EQ delivers estimates of relevant ground motion parameters, such as peak ground velocity, peak ground acceleration, or shaking duration, with very high spatial resolution. The physics-based engine includes pre-trained Machine Learning (ML) models fed with pre-computed simulation databases, as well as deterministic 3D simulations on demand, providing results in minutes and hours, respectively. The combined results, when well-calibrated, could lead to a new generation of ground shaking maps that complement GMPEs for rapid hazard assessment.To demonstrate the potential use of UC in seismology,  in this work we show the UCIS4EQ simulation of the M7.1 Puebla earthquake that occurred in central Mexico on the 19th of September 2017. With a hypocentre at 18.40ºN, 98.72ºW and 57 km depth, the Puebla earthquake was located about 150 km southeast from Mexico City. Identified as a severe event (VIII) in the Modified Mercalli Intensity scale, it resulted in a total of 370 killed and around 6000 injured, as well as structural damages, downed telephone lines, and ruptured gas mains.

Pre-print: Ground Motion Shaking Predictions Based on Machine Learning and Physics-based SimulationsEarthquakes constitute a major threat to human lives and infrastructure, hence it is crucial to quickly assess the intensity of ground motions after a major seismic event. Rapid estimation of the intensity of ground vibrations is essential to assess the impact after a major earthquake occurs. The Machine Learning Estimator for Ground Shaking Maps (MLESmap) introduces an innovative approach that harnesses the predictive capabilities of Machine Learning (ML) algorithms, utilizing high-quality physics-based seismic scenarios. MLESmap aims to provide ground intensity measures within seconds following an earthquake. The inferred information can produce shaking maps of the ground providing quasi-real-time affectation information to help us explore uncertainties quickly and reliably. To develop the MLESmap technology, we used ground-motion simulations generated by the CyberShake platform. Originally designed for Southern California, this physics-based Probabilistic Seismic Hazard Methodology was migrated to the South Iceland Seismic Zone recently. Our methodology follows a three-step process: simulation, training, and deployment. By employing this approach, we can generate the next generation of ground shake maps, incorporating essential physical information derived from wave propagation, such as directivity, topography, and site effects. Remarkably, the evaluation times for MLESmap are comparable to empirical Ground Motion Models, whereas the predictive capacity of the former is superior for the Mw > 5 earthquakes.In this work, we present the application of the MLESmap methodology in South West Iceland.