Abstract: Wireless sensor network protocol design is highly depended on the link characteristics of the network environment. In real application field, link reliability is vulnerable to environment interferences. But there are little works which reveal the temporal variation and multi-channel difference characteristics of the link reliability. The authors present a statistics approach according to the temporal and frequency characteristics of link reliability in industrial environment. Two variables, i.e. packet drop rate (PDR) and packet drop sequence (PDS) are introduced to describe the relationship of link reliability with time and frequency. Through the measurements in typical environments like iron and steel factory and experimental factory in USTB with IEEE 802.15.4 compliant transceivers, it is found that the PDR of a certain range follows logistic distribution or log-logistic distribution, and the interval between two dropped packets for a given PDR follows lognormal distribution as well. Furthermore, based on the analysis of the link reliability, a link layer statistical model is presented and two proposals for IEEE802.15.4e are evaluated on the new model. The simulation results show that the network reliability can be improved when taking the statistic characteristics of the link variations into consideration in the design of high reliable protocols.