CONCYTEC-WORLD BANK PROJECT: MACHINE LEARNING AND EARTH OBSERVATION TECHNOLOGIES FOR DISASTER MITIGATION

Moya L.; Gonzales C.; Diaz M.; Adriano B.; Mas E.; Zavala C.; Koshimura S.; Yamazaki F. · World Conference on Earthquake Engineering proceedings (2021)

Resumen

Currently, there are several types of sensors that are used for disaster science. Accelerometers for ground motion recording, tidal gauges for tsunami monitoring, and satellite images for change detection are such examples. Field surveys might also be considered as a form of earth observation from which essential products like fragility functions have been obtained. This paper summarizes the research framework of the project "Fusion of Machine Learning Algorithms and Earth Observation Technologies for Disaster Mitigation," which aims to merge all the referred sources of information to identify the affected area in the aftermath of a large-scale disaster in a fully automatic mode. For this purpose, in-place sensors that measure the intensity of the disaster in a given coordinate are used to calibrate a disaster's numerical model. The project focuses on earthquakes, and thus, ground motion networks are used to compute the peak ground acceleration (PGA), the peak ground velocity (PGV), and the seismic intensity at the station's coordinates. Then, a numerical interpolation method that considers attenuation laws is used to compute the referred seismic parameters in a uniform spatial grid, referred to as an intensity map. The intensity map is used with fragility functions to obtain prior information of the expected damage at an arbitrary coordinate. Finally, the prior information is used with satellite images to allocate real damaged areas. The fusion of the intensity map, the fragility functions, and the satellite imagery is performed using machine learning techniques. It is worth noting that the machine learning algorithm was adapted to use intensity maps and fragility functions instead of training data, which enables its application in a fully automatic form. The project is conducted within the framework and financial support of the Concytec - World Bank "Improvement and Extension of the Services of the National System of Science, Technology and Technological Innovation (SINACYT)" Project 8682-PE, through its executive unit Fondecyt (contract 038-2019).

Instituciones: Universidad Nacional de Ingeniería