This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as

This paper aims to disclose the law of fish migration trajectories at different water depths. For this purpose, the grass carps in a reservoir in southwestern China were taken as the targets, outdoor experiments were performed to monitor their behaviours and environmental factors in the reservoir. Then, the Hydroacoustic Technology, Inc. (HTI) acoustic tracking system and backpropagation neural network (BPNN) were introduced to simulate and analyse the migration of the fish in the natural state. Meanwhile, the vertical distribution of fish was discussed at different temperatures and dissolved oxygen contents. The results show that the BPNN algorithm has a good fitting effect on the planar migration trajectories of the fish, but fails to achieve a desirable fitting result concerning the migration trajectories in the Z direction. Fortunately, the fitting effect of migration trajectories was greatly enhanced by normalization. The fish were distributed differently in spring and summer across the different water depths, under the influence of water temperature and dissolved oxygen content. Overall, the fish obeyed the normal distribution in the vertical direction, and selected water depth mainly based on dissolved oxygen content. The research findings lay a scientific basis for fish resource protection, river ecology assessment and water environment restoration.