Unexpected events, such as crashes, disabled vehicles and spilled loads cause traffic congestion or increase the level of existing congestion on roads. Traffic incident management requires the estimation of incident duration from the beginning to the final stage of the event, in order to identify the most suitable strategies for decreasing their impact. Accordingly, the adverse effects of an incident can be minimised by efficiently managing the process between the time the incident occurred and the traffic returning to normal operating conditions. The purpose of this paper is to present a methodology and improve the accuracy of the model by using a multi-step approach to estimate the duration of an incident. For this purpose, incident duration data was obtained for Istanbul Transit Europe Motorways (TEM). Using the incident data and implementing the multi-step approach, Kalman Filtering and linear regression analysis in stepwise, were used for the incident duration estimation model.