Machine Learning Forecasting of Romanian Earthquakes: Challenges and Opportunities

Romania is a country situated in Eastern Europe which encompasses an important seismic region, the Vrancea region. Large earthquakes occurring in this region have the potential of producing a massive amount of casualties and damage in Romania and the neighbouring countries. For this reason, preparation for future earthquakes is a priority. With the advent of Machine Learning, seismicity analysis and forecasting entered a new era. Robust methods such as those based on neural networks have been employed for forecasting earthquakes in various regions around the Globe. The present paper reunites a number of such approaches, focusing on those that derive their information from historic magnitude sequences. Common data preprocessing and learning techniques are described. With respect to Romania, highlighted studies focus on earthquake impact mitigation through risk assessment and the development of early warning systems as well as the incipient use of machine learning modeling. We systematize the current challenges and opportunities and propose guidelines for future Machine Learning exploration of Romanian seismicity, with a focus on data available in public magnitude catalogs and neural networks. To the best of our knowledge, this is the first comprehensive study that pursues the perspective of developing an intelligent earthquake forecasting system for Romania considering data derived from historic magnitude sequences and neural networks.