Meteorologically Driven Neutron Background Prediction for Homeland Security

This paper uses the meteorological conditions to predict neutron background counts as measured by a neutron detector. Such counts occur due to cosmic-ray interactions and temporal variations that are primarily dependent on meteorological factors. This paper presents a marginal autoregressive (AR) model constructed from eight months of hourly meteorological data that accurately predicts neutron background without being susceptible to the presence of source material. Nonmeteorological factors impacting neutron background (including cosmic-ray intensity) are not readily measurable in real time in a validated format, so a conditional AR model is also presented that can compensate for phenomena not included in the model while maintaining some independence from potential sources. This paper shows that the conditional AR model is more effective than the moving average, with lower susceptibility to source contamination, ensuring a consistent level of sensor effectiveness.