Forecasting Strong Subsequent Aftershocks in New Zealand: Preliminary Results

NESTORE (NExt STRong Related Earthquake; Gentili et al., 2023) is an algorithm for the probabilistic forecasting of strong aftershocks after a major seismic event. The algorithm evaluates nine features that characterize the seismic activity at increasing time intervals after the mainshock. These features are then analyzed using a combination of supervised machine learning and statistical validation to estimate the probability that an initial strong event of magnitude Mm will be followed by another event of magnitude ≥ Mm–1 within a given space-time window associated with the corresponding seismic cluster. If such an aftershock occurs, the cluster is classified as »Type A« (indicating a higher potential risk). The algorithm outputs the probability that the cluster is of type A. New Zealand is one of the most seismically active regions in the world, located on the boundary between the Australian and Pacific tectonic plates. The complex country’s seismicity has generated so far earthquakes up to magnitude 7.8. Understanding and forecasting seismic activity is therefore critical for risk assessment and mitigation efforts. We present here the preliminary results of the applications of NESTORE to the seismicity of New Zealand. We split the dataset of the clusters in the area between training (1988–2015) and testing (2016–2025) for a retrospective forecasting. We refined the training dataset using the outlier detection method REPENESE (RElevant features, PErcentage class weighting, NEighborhood detection and SElection), which was developed for skewed distributions of feature values (Gentili et al., 2025). We found that twelve hours after the first earthquake 88 % of the clusters were correctly classified. Funded by the RETURN project European Union Next-GenerationEU (National Recovery and Resilience Plan, PE0000005).