Abstract: The event and event boundary detection is one of the most important applications of wireless sensor networks(WSNs) . The accurate node fault detection is the premise of improving the efficiency of the event and event boundary detection. However, current approaches consider fewer fault types of sensor nodes, which are very possible to mistake the event boundary nodes for the fault nodes; and current approaches also need nodes to communicate with each other frequently. Thus current approaches usually lead to low level of the fault-tolerant ability of WSNs systems, bad availability of nodes and high energy consumption. Therefore, in order to increase the accuracy of detection rate and the efficiency of energy consumption, a new fault-tolerant event and event boundary detection algorithm for wireless sensor networks is proposed. The algorithm adopts the temporal correlated information to detect event, and adopts the spatial correlated information to detect fault and event boundary. The node reliability level restore mechanism is also proposed, enabling nodes to act according to the change of network environments, and adjust their reliability level automatically. Experimental results show that the fault-tolerant event and event boundary detection algorithm proposed in this paper has good performance even in the conditions of high fault probabilities.