Abstract: Real-time estimation and diagnosis of slope early warning indicators are important means to monitor the safety state of slope and prevent the situation of the evolution from slope to landslide. Safety early warning pays more attention to extreme events, which are not common. The key to achieve safety early warning lies in whether we can accurately analyze the quantile of the slope effect quantity sequences and describe the tail characteristics of the slope effect quantity sequences. First of all, the slope displacement monitoring sequence can be divided by the interval of time; then, the thresholds of the slope displacement monitoring sequences at each time interval are drafted using Hill figure method; the part of the displacement monitoring sequence at each time interval that is supra-threshold is fitted by Generalized Pareto Distribution (GPD), and the tail characteristics of the part that is supra-threshold of displacement monitoring sequences are described progressively. So, the ability of the slope to resist historical loads is analyzed based on the model of peaks over threshold (POT) using extreme value theory. If the failure probability of the slope is provided, the real-time estimation sequences of early warming indicators of displacement can be obtained and the ability of the slope to resist extreme loads that may happen in the future can be exploited and evaluated. Finally, the real-time estimation sequences of early warming indicators of displacement are diagnosed by using cusp catastrophic model from catastrophe theory. Based on specific slope projects, the rationality of the mixed model of POT and catastrophe theory is verified and the modeling method of the mixed model will be a new way to achieve the real-time estimation and diagnosis of engineering early warning indicators.