Traffic operations vary drastically before-, during- and after natural disasters due to several reasons, such as the movement of affected populations, road failures, and heavy precipitation. The disaster-induced effects on road links and their subsequent recovery to the pre-disaster state are functions of the designed network resilience. Thus, the investigation of traffic operations including speed and volume fluctuations during such disasters can provide useful insights into the resilience of the roadway network. Furthermore, identifying road links and corridors which are most affected during natural disasters and estimating the extent of the effect on traffic are crucial steps in devising traffic management strategies for mitigating future hazards. A key challenge to the realization of the above objectives is the lack of data pertaining to roadway and traffic conditions during natural disasters. While modern sensing technologies based on non-Dedicated Short-Range (wireless) Communications such as Bluetooth® have been widely adopted to track traffic performance, they need to be corroborated with supplementary data for a complete characterization of the prevalent traffic conditions. In this study, an alternative method is presented for identifying the traffic fluctuations induced by a natural disaster (Hurricane Harvey) on an urban traffic network (Houston) by studying the characteristics of extreme travel time observations. The study relies on time series decomposition and anomaly detection algorithms to investigate the spatiotemporal effects of the hurricane on the traffic conditions. The results of the case study suggest that the metrics developed are effective in quantifying the resilience of traffic networks against natural disasters by capturing both the initial impact and recovery. The study showed that the Houston freeway network was not completely recovered from the effects of Hurricane Harvey, even three weeks after the occurrence of the hurricane.@en