In order to clarify how different environmental sample types affect the detectability of mussel species when using environmental DNA metabarcoding technology, surface water, bottom water and sediment were collected in Poyang Lake in winter and spring of 2021 for environmental DNA metabarcoding analysis, and then combined with traditional methods for collection and verification. A total of 33 species of mussels from Poyang Lake were detected by environmental DNA metabarcoding technology, while 18 species were collected by traditional methods. All the species collected by traditional methods could be detected by environmental DNA metabarcoding technology. The number of mussel species annotated in surface water and bottom water was respectively higher than that in sediment, and the mussel species annotated in surface water and bottom water completely covered sediment, respectively. There was no significant seasonal difference in α diversity level of mussels based on environmental DNA metabarcoding, but significant seasonal difference in β diversity level of mussels. The mussel diversity in both surface and bottom water was significantly respectively higher than that in sediment samples. The Beta diversity analysis also showed significant differences between water samples (surface and bottom water respectively) and sediment samples. But there were no significant differences in the diversity and community structure between surface and bottom water. Water depth (WD), depth of Secchi disk (SD), water temperature (WT) and total nitrogen (TN) significantly affected the community structure of mussels in Poyang Lake. Environmental DNA metabarcoding is feasible in monitoring mussel freshwater diversity, and water sampling is better than sediment sampling. There is no significant difference between surface water and bottom water.