Transmission lines (TL) are a very important part of our infrastructure. Their design is still mainly based on a single extreme value wind speed, evaluated from synoptic or mixed wind speed records, whereas non-synoptic (e.g. tornado and downbursts) winds are responsible for up to 80% of weather-related TL collapses. In this manuscript a methodology is proposed to evaluate the reliability of complete TL segments, considering the large uncertainties in wind speeds, tornado diameter and relative trajectory w.r.t. TL supports (tower offset). The Performance-Based Wind Engineering framework is employed to account for the uncertainties in wind speeds, tornado diameter and tower offset. A compact non-linear dynamic scheme is employed to handle the material and geometric non-linearities of a tower-cable TL segment, capturing the influence of cables in the dynamic response of the TL. Mean wind profiles and turbulent velocity field models are employed for simulating tornado loading in time domain. Fragility analysis is carried out for three performance levels (Serviceability, Damage control and Collapse). Results show that TL collapse is conditional on tower-hit events. The probability of a tower hit event is determined from geometrical relationships between tower span, tornado radius and tornado trajectory. The probability of a tower hit event increases significantly with tornado radius. Yet, uncertainty in tornado radius is found to be less relevant to TL vulnerability than uncertainty in wind speeds and tower offset.