Neural Negative Binomial Regression for Weekly Seismicity Forecasting: Per-Cell Dispersion Estimation and Tail Risk Assessment
arXiv:2605.21437v1 Announce Type: cross Abstract: Standard approaches to forecasting the weekly number of earthquakes on a spatial grid rely on the Poisson distribution with a single global dispersion assumption. We show that this assumption is systematically violated in seismic data from Central Asia (2010-2024), where a likelihood-ratio test with boundary correction strongly rejects the Poisson hypothesis (p = 5), where the continuous ranked probability score (CRPS) of the proposed model is 12.5 percent lower than that of the baseline, indicating improved calibration in extreme-event forecas
The increasing availability of large seismic datasets and advances in neural network architectures are enabling more sophisticated approaches to geophysical modeling.
Improved seismic forecasting models can lead to more accurate risk assessments for critical infrastructure and urban planning in earthquake-prone regions.
The ability to accurately model and forecast extreme seismic events is enhanced by moving beyond simplistic statistical assumptions and incorporating more complex, data-driven approaches.
- · Geophysicists
- · Insurance industry
- · Civil engineering
- · Disaster preparedness organizations
- · Outdated seismic forecasting models
- · Regions lacking advanced computational resources
More precise weekly earthquake forecasts, especially for extreme events, can be generated through advanced AI models.
Better risk mitigation strategies and infrastructure planning can be developed based on these improved forecasts, potentially reducing economic losses and casualties.
The application of similar AI methodologies could extend to forecasting other complex natural phenomena, impacting fields from meteorology to climate science.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG