Overview of portal monitoring at border crossings

TLDR

CBP must interdict illicit radioactive material at ports of entry, targeting RDDs, nuclear warheads, and SNM while maintaining high detection probability in primary screening without flooding secondary checks with alarms from naturally occurring radioactive material. The primary survey screens all vehicles, diverting suspicious ones to secondary, using energy‑based alarm algorithms that account for portal baseline depression and leverage plastic scintillator energy data to differentiate NORM from SNM, with performance evaluated on a large empirical dataset to set false‑alarm thresholds and estimate detection probabilities for marginal SNM sources. Energy‑based algorithms yield significantly higher detection probabilities for small SNM sources than gross‑count algorithms when NORM cargo limits the alarm threshold.

Abstract

The Bureau of Customs and Border Protection has the task of interdicting illicit radioactive material at ports of entry. Items of concern include radiation dispersal devices (RDD), nuclear warheads, and special nuclear material (SNM). The preferred survey method screens all vehicles in primary and diverts questionable vehicles to secondary. This requires high detection probability in primary while not overwhelming secondary with alarms, which could include naturally occurring radioactive material (NORM) found in acceptable cargo and radionuclides used in medical procedures. Sensitive alarm algorithms must accommodate the baseline depression observed whenever a vehicle enters the portal. Energy-based algorithms can effectively use the crude energy information available from a plastic scintillator to distinguish NORM from SNM. Whenever NORM cargo limits the alarm threshold, energy-based algorithms produce significantly better detection probabilities for small SNM sources than gross-count algorithms. Algorithms can be best evaluated using a large empirical data set to 1) calculate false alarm probabilities, 2) select sigma-level thresholds for operationally acceptable false alarm rates, and 3) determine detection probabilities for marginally detectable pseudo sources of SNM.