MLESmap is an artificial intelligence-powered system that creates near-real-time maps of earthquake impacts to help with emergency response and disaster management. The project provides quick and accurate maps and graphs on using high-performance computing, physics-based models, and machine learning.

Problem

Earthquakes are one of the most dangerous natural disasters because they are unpredictable. In the minutes following an earthquake, emergency services and civil protection organizations require quick and precise information to determine which areas are most damaged. However, early assessments of earthquake consequences are sometimes unclear or slow, reducing the effectiveness of response activities in the vital early phases.

Solution

MLESmap addresses this challenge by running a large number of realistic, physics-based earthquake simulations ahead of time using high-performance computing. These simulations cover a wide range of earthquake scenarios and are used to train machine learning models that understand the relationship between earthquake characteristics and ground motion. When an actual earthquake happens, the system uses basic event information, such as location and magnitude, to quickly estimate ground acceleration using pre-trained models. This enables decision-makers to receive accurate estimates of shaking strength in near real time, as well as clear maps and visual summaries.