Sismólogos avanzan en un modelo de aprendizaje profundo para predecir terremotos - INVDES
Researchers at the University of California have developed RECAST, a deep learning model that significantly improves the prediction of large-magnitude earthquake aftershocks. By analyzing vast historical and current seismic datasets, this approach outperforms the traditional ETAS method, which was designed for an era of limited data. RECAST demonstrates superior accuracy and speed, proving that modern computational capabilities can handle the complexity of extensive seismic catalogs more effectively than previous statistical techniques. The relevance to open data lies in the model’s ability to leverage comprehensive, continuously recorded ground movement information rather than relying on rigid definitions of specific events. This shift encourages the use of all available seismic data, fostering a more inclusive and robust approach to analysis. The flexibility of RECAST allows it to be trained on diverse global datasets and adapted to regions with fewer resources, highlighting the potential of open, standardized data sharing to enhance predictive capabilities worldwide. Ultimately, this advancement signifies a methodological shift in seismology, where the volume and quality of open data directly drive scientific progress. It underscores the importance of accessible, large-scale datasets in developing reliable forecasting tools. By moving beyond fragile traditional models, researchers can create adaptable systems that utilize continuous monitoring information, improving disaster preparedness and scientific understanding through the effective integration of open seismic records.
Source: invdes.com.mxPublished on 2023-09-20