Fatigue has gradually become a serious issue for orthotropic steel deck used for long-span bridges. Two fatigue effects, namely number of stress cycles and equivalent stress amplitude, were introduced as investigated parameters in this paper. Investigation was focused on their relationships with traffic volume and ambient temperature by using 7-months fatigue monitoring data of an actual bridge. A fatigue analytical model considering temperature-induced changes in material property of asphalt pavement was established for verifying these relationships. The analysis results revealed that the number of stress cycles and equivalent stress amplitude showed a linear correlation with the traffic volume and ambient temperature, respectively, and that the rib-to-deck welded joint was much more sensitive to the traffic volume and ambient temperature than the rib-to-rib welded joint. The applicability of the code-recommended model for fatigue vehicle loading was also discussed, which revealed that the deterministic vehicle loading model requires improvement to account for significant randomness of the actual traffic conditions. Fatigue has gradually has a serious issue for orthotropic steel deck used for long-span bridges. Two fatigue effects, of number of stress cycles and equivalent stress amplitude, were introduced as investigated parameters in this paper. Investigation was focused on their relationships with traffic volume and ambient temperature by using 7-months fatigue monitoring data of an actual bridge. A fatigue analytical model considering temperature-induced changes in material property of asphalt pavement was established for verifying these relationships. The analysis results revealed that the number of stress cycles and equivalent stress amplitude showed a linear correlation with the traffic volume and ambient temperature, respectively, and that the rib-to-deck welded joint was much more sensitive to the traffic volume and ambient temperature than the rib-to-rib welded joint. the code-recommended model for fatigue vehicle loading was also discussed, which revealed tha t the deterministic vehicle loading model requires improvement to account for significant randomness of the actual traffic conditions.