The Internet of Things (IoT) technology enables physical devices to communicate with each other for preparing, gathering, and sharing hazard warnings or critical information without human intervention. With respect to emergency applications of IoT technology, an essential issue is to provide an efficient and robust scheme for data gathering. The proposed solution in the existing approaches is to construct a spanning tree over the IoT devices and collect data using the tree. The shortcoming of these algorithms is that they do not take into account the probability of device mobility or failure. In such cases, the spanning tree is split, and it becomes impossible to deliver critical data to the base station on time. In this paper, we propose a reliable spanning tree construction algorithm, which is called reliable spanning tree construction in IoT (RST-IoT). Our algorithm utilizes the artificial bee colony algorithm to generate proper trees. In this method, hop count distances of the devices from the base station, residual energies of the devices, and their mobility probabilities are considered to measure the appropriateness of the trees. Moreover, the proposed algorithm generates a number of trees instead of a single one. These trees are arranged according to their preferences and used for data gathering in succession. Each tree is employed for data gathering upon splitting the preceding one. The simulation results show that RST-IoT improves the reliability of data gathering in emergency applications compared to the previous approaches.