One of the most important challenges of distributed fiber-optic intrusion detection systems is to minimize the nuisance alarm rate without compromising the probability of detection in a wide range of operating environments. This involves eliminating nuisance alarms caused by non-intrusion events such as torrential rain without compromising their sensitivity to intrusion events. An effective yet computationally non-intensive event recognition and discrimination technique is presented for eliminating rain-induced nuisance alarms. Results from real intrusion detection systems are presented showing the elimination of rain-induced nuisance alarms for torrential rainfall rates in excess of 4 inches/hr without any penalty to the simultaneous detection sensitivity of intrusion events.