||Travel time reliability, connectivity reliability and level of service are always regarded as an indicator to measure the performance of a network and there are numerous ways to evaluate network performance under normal conditions. However, few of them concentrate on considering the network performance subjected to the non-recurrent events such as earthquakes, floods, traffic accidents, and adverse weather which would induce serious traffic jams and lead to an uncountable loss of economy of a city. Most researches investigated how to reduce average delay in this situation; however, the consequences of non-recurrent congestion in a network were rarely mentioned by other literature. This thesis aims to fill this gap and discusses the network performance under the impact of non-recurrent congestion in a simple network. Traffic information is usually thought to be a solution to minimize the deterioration of network performance when the non-recurrent congestions happen. In this thesis, simulations are carried out in a dynamic parallel network and a general picture of consequences of non-recurrent network capacity degradation is captured. It also provides a mechanism to assess the network performance under consideration of uncertainties such as adverse weather, traffic accident and natural disasters. Among natural disasters, slope failures are particularly prevalent in Hong Kong. An actual road network in Hong Kong is selected and the cost benefit analysis is carried out to demonstrate the trade-off between the investment and traffic information to improve the network performance.