Because of its direct influence on the amount of unfrozen water and ice lens in a frozen soil, temperature has a significant effect on the mechanical behavior of the freeze-thaw soil. Accordingly, seismic responses of engineering structures such as canals in northern cold regions exhibit noticeable differences with seasonal alternation. To analyze the distinctive seismic characteristics of a canal in northern cold regions, a coupled water-heat-dynamics model was built based on theories of heat transfer, soil moisture dynamics, frozen soil mechanics, and soil dynamics. A well-monitored canal in northern cold regions was used to simulate seismic responses in two typical seasons in the 10th service year. The numerical results showed that the constructed canal disturbed the original thermal state and the geo-temperatures of the canal changed with seasonal alternation. In the freezing-thawing process, the unfrozen water migrated and ice lens formed in the canal soil under the forcing of temperature gradient. As a result, the unfrozen water and ice distributions of the canal exhibited obvious seasonal differences. For instance, there were little unfrozen water and much ice lens in the canal and shallow layer of soil at air temperature bellowing freezing, whereas the volumetric content of the unfrozen water was high and the ice content was equal to 0 when the temperature had its maximum during the year (on July 15). So the differential seismic responses of the canal were generated by different water-heat states in the two seasons. Among these two seasons in the 10th service year, although the general time histories of the acceleration and velocity were similar in the canal, their maximum amplitudes were relatively larger on July 15. For instance, in two different time (on January 15, on July 15), the accelerations of the canal bottom respectively got their maximum acceleration values which were 1.160 m/s2, 1.360 m/s2 in 12.165s and 2.404 s, whereas the accelerations of the canal top got their respectively in 8.995 s and 9.007 s maximum values which were 1.476 m/s2, 1.785 m/s2, so the earthquake accelerations of the canal were stronger on July 15 than January 15 under the earthquake. The horizontal velocity responses of the bottom and the top of the canal were smaller on January 15, their maximum velocity was 0.145 m/s and 0.149 m/s, respectively, while the maximum speed was 0.146 m/s and 0.150 m/s, respectively on July 15. The displacement was a direct expression of the seismic loading and also sensitive to temperature variation, the displacement responses of the canal to dynamic loading are also relatively larger on July 15 and the maximum horizontal displacement was 5.6 cm in the example. When the earthquake was over, there were still permanent differential deformations in the canal, and the residual displacement distributions of the canal were asymmetrical. In fact, the seismic response of the canal in seasonally frozen soil area was a very complicated water-heat-dynamic coupled problem, so there were few theoretical documents on the subject up to now. As a preliminary study, the numerical model proposed in this paper has some limitations. Therefore, further studies should be carried out on the subject. As a preliminary exploration, it is expected to provide theoretical basis and reference for design, construction, and maintenance of the canal in seasonally frozen regions. [ABSTRACT FROM AUTHOR]