Several prediction models are checked against experimental freezing rain si
mulations and field observations. Goodwin's model is used to estimate ice l
oads occurring in Winnipeg, Canada, from the available 126 years of weather
data. Emphasis is put on predicting extreme ice loads from this historical
record. A Monte Carlo simulation is proposed to address the random variati
ons of ice storms. The simulation suggests that the current practice of usi
ng a return period-based, extreme ice load as the design ice load is unsafe
unless the return period is substantially longer than a power line's lifet
ime. A probability-based approach is suggested as a better alternative.