Descripción del proyecto
Detección de señales de baja amplitud para anticipar seísmos
Los riesgos naturales, como los seísmos, son difíciles de predecir. Sin embargo, los impresionantes avances en el ámbito de la inteligencia artificial (IA) están allanando el camino para la anticipación de fenómenos destructivos. El proyecto financiado con fondos europeos EARLI utilizará la IA para identificar señales sísmicas débiles y tempranas que permitan no solo agilizar una alerta rápida, sino también explorar la posibilidad de predecir seísmos. En particular, aplicará un método de alerta rápida basado en una nueva señal identificada, provocada por la perturbación del campo de gravedad generado por un seísmo, que es alrededor de seis órdenes de magnitud más reducido que las ondas sísmicas (lo cual limita enormemente su detección con técnicas estándar), pero las precede. El segundo objetivo, de carácter más exploratorio, será adaptar el algoritmo de IA desarrollado para encontrar señales incluso más tempranas que precedan el surgimiento de seísmos de gran tamaño.
Objetivo
Earthquakes caused nearly one million fatalities in the last two decades. The hazardous nature of earthquakes is largely due to their unpredictability. The question of whether this unpredictability is ontological (i.e. earthquakes are a chaotic phenomenon that physics cannot predict) or a consequence of our incapacity to model them is still open. In the first case, one may never hope to predict earthquakes and efforts should be focused towards developing early-warning approaches so that the population can prepare for imminent shaking and tsunami. In the second case, earthquake prediction becomes theoretically achievable. In both cases, Artificial Intelligence (AI) may lead to giant steps in anticipating destructive events. I propose here to use AI to identify weak early seismic signals to both speed up early-warning and explore the possibility of earthquake prediction. The first part of the project will be devoted to implementing an early-warning approach not based on P-waves as all current systems but on an earlier signal recently identified. This signal is due to the perturbation of the gravity field generated by an earthquake – which propagates at the speed of light – but is ~6 orders of magnitude smaller than seismic waves, strongly limiting its detection with standard techniques. AI has proven very efficient at detecting low-amplitude signals. I will implement an AI algorithm to systematically detect gravity perturbations generated by magnitude > 7 earthquakes and rapidly estimate from them the location and magnitude of the earthquake. Though the existence of earthquake precursors (i.e. signals preceding the origin of earthquakes themselves) is hypothetical, AI represents a new prowerful mean to discover them. In the second part of the project, I will adapt the AI algorithm developed in the first part to search for earthquake precursors.
Ámbito científico
Palabras clave
Programa(s)
Régimen de financiación
ERC-STG - Starting GrantInstitución de acogida
13572 Marseille
Francia