Nuisance alarms are distinct from false alarms in the strict sense (alarms with no cause at all), though the terms are often used interchangeably. NAR matters because operator assessment capacity is finite: in our fusion model one operator can assess about 20 alerts per hour, and a 40-zone OR-fused perimeter produced 76 per hour on a windy night. Coincidence (AND) fusion reduces chance alarms to about per zone but multiplies detection probabilities and assumes independent nuisance sources.
Measuring NAR
Record every alarm during a baseline period and classify its cause after assessment. Report NAR per zone per day, split by weather, because a single average hides the nights that matter. A zone whose NAR is ten times its neighbours' is usually a placement problem — vegetation, a road, drainage — rather than a sensor fault.
Reducing NAR without losing detection
Fuse sensors with physically different failure modes; verify alerts visually before they reach an operator; and fix the placement problems that baselining reveals before turning sensitivity down.