The accuracy of bridge system safety evaluations and reliability assessments obtained through refined structural analysis procedures depends on the proper modeling of traffic load effects. In fact, traffic load intensities and related uncertainties need to be defined based on available real traffic data supplemented by a statistical model able to project the data in time to estimate the maximum load effect expected over a specific bridge design or service period. This paper proposes a procedure to calibrate appropriate live load models that can be used for advanced analyses of bridges. The calibration procedure is demonstrated using actual truck data collected at a representative weigh-in-motion (WIM) station in New York State. Extreme value theory is used to project traffic load effect to different service periods using the upper tail of the load ef-fect probability distribution. The results are presented in live load models developed for a 5-year typical rat-ing interval and for a 75-year design life. Live load models obtained using the proposed calibration procedure are readily implementable for deterministic refined analyses of highway bridges and for evaluating the relia-bility of bridges at serviceability and ultimate limit states considering the system’s behavior. Maximum traf-fic load is calculated with the help of a two-truck configuration which is the governing load case in many short and medium-span bridge configurations. The results are presented in terms of standard AASHTO 3-S2 Legal Rating trucks, with appropriate reference-period-dependent weight adjustment factors.