Through comparing the measured data of dynamic strains due to loading and temperature by the stain gauge and temperature sensor at the same location, the information in the strain data was divided into three parts in the frequency domain by using the defined index named power spectral density (PSD)-ratio index. The three parts are dominated respectively by temperature varying, stresses, and noises and thus can be distinguished from the determined the separatrix frequencies. Also, a simple algorithm was developed to separate the three types of information and to extract the strain caused mainly by structural stresses. As an application of the proposed method, the effect of strain deformation and noises on the fatigue assessment was investigated based on the separated data. The results show that, the determined values of separatrix frequencies are valuable for the monitoring data from other bridges. The algorithm is a multiresolution and hierarchical method, which has been validated as a simple and effective method for data analyses, and is suitable for the compression and preprocessing of the great amount monitoring data and easy to be integrated into the structural health monitoring (SHM) soft system. The strain due to temperature varying attributes a little to the errors of fatigue assessment; however, the noises or random disturbance existed in the monitoring data have much responsibility for the errors, and the main reason is that the random disturbance shifts the real strain/stress amplitude picked up by real structural stress or strain.