Temporary on-ramp closure has been proposed as a strategy to reduce the impact of severe incidents on freeway facilities; however, to date no rigorous procedure has been made available to provide guidance on how such a technique should best be used. In particular, one must decide which ramps to close and for how long. A two-phase approach is proposed to answer these questions. The first phase is macroscopic in nature and predicts how motorists will reroute in response to any ramp closure and recommends which ramps should be closed. The second phase uses microsimulation to study the vicinity of the incident in greater detail, more fully accounting for dynamic traffic phenomena and attempting to answer the question of how long these ramps should be closed. From a computational standpoint, the first phase is designed to run as quickly as possible to allow the ramp closure policy to be enacted as the second phase begins, since the results of the second phase are not needed until later. This procedure is demonstrated by using a fictitious incident in the El Paso region of Texas.