Dams are built to supply water to users and often to protect people and properties against floods in downstream areas. Efficiency of dams for flood control is improved substantially if a flood forecasting system is implemented. Rainfall threshold (RT) depths correspond to the occurrence of critical discharge at given cross-sections for given rainfall durations and initial soil moisture conditions of the upstream watershed. Here, we present an RT-based approach for offline flood forecasting downstream of dams. The proposed methodology incorporates rainfall-runoff and reservoir routing models while the spatial distribution of rainfall is probabilistically modeled based on a Monte Carlo approach. The RT curves are derived as a function of initial water elevation in the reservoir. The algorithm is implemented for a flood-prone area downstream of a dam in southwestern Iran. The results showed a clear rise in the RT values compared to the no-dam case, which is mainly due to the reservoir routing effect. The rate of rise in the RT values decreased with higher initial water elevation in the reservoir. The proposed method also provides the operator with the flexibility of adopting one of the various RT curves subject to different probabilities based on risk tolerance.