Bayesian Event Categorization Matrix Approach for Explosion Monitoring

Current efforts to correctly categorize natural events from suspected explosion sources with data that is collected by ground- or space-based sensors presents historical challenges that remain unaddressed by the Event Categorization Matrix (ECM) model. Smaller historical events (lower yield explosions) often include only sparse observations among few modalities and can therefore lack a complete set of discriminants. The covariance structures can also vary significantly between such observations of event (source-type) categories. Both obstacles are problematic for the ``classic'' Event Categorization Matrix model. Our work addresses this gap and presents a Bayesian update to the previous Event Categorization Matrix model, termed the Bayesian Event Categorization Matrix model, which can be trained on partial observations and does not rely on a pooled covariance structure. We further augment the Event Categorization Matrix model with Bayesian Decision Theory so that false negative or false positive rates of an event categorization can be reduced in an intuitive manner. To demonstrate improved categorization rates for the Bayesian Event Categorization Matrix model, we compare an array of Bayesian and classic models with multiple performance metrics using Monte Carlo experiments. We use both synthetic and real data. Our Bayesian models show consistent gains in overall accuracy and a lower false negative rates relative to the classic Event Categorization Matrix model. We propose future avenues to improve Bayesian Event Categorization Matrix models for further improving decision-making and predictive capability.