A method for determining alarm thresholds based on noise data characteristics is proposed, and this threshold is applied to analyze the potential alarm distribution of single/total strain monitoring data on temporal and spatial scales. Firstly, simulation tests are conducted to verify the effectiveness of wavelet denoising. Then, distributed optical fibers with a length of kilometer are laid at the tunnel for strain monitoring. Wavelet transform is applied in denoising analysis of distributed fiber-optic monitoring data in the tunnel. The results show that wavelet denoising can effectively restore effective signals. The warning threshold determined based on the characteristics of noise data can identify risk monitoring points more efficiently and reduce the probability of invalid alarms to a large extent. Moreover, the two locations of tunnel damage discovered through manual investigation are consistent with the risk points obtained based on fiber optic monitoring data analysis, partially verifying the accuracy of the risk points. At the same time, combined with the analysis results of the monitoring data from the tunnel and slope borehole inclinometers, it is found that the tunnel and the existing slopes have overall continuous slow creep and local sudden changes, which are risky. Strengthening monitoring and taking further preventive measures to ensure the safety of line operations are suggested. This research provides further reference for the application of wavelet denoising to extract effective data from massive data in engineering monitoring and the application of distributed optical fiber in deformation monitoring and early warning of various types of rock and soil engineering.