AbstractAs fish communities are a major concern in rivers ecosystems, we investigated if their environmental (e)DNA signals vary according to the sampling period or hydromorphological conditions. Three rivers were studied over a year using eDNA metabarcoding approach. The majority of the species (c. 80%) were detected all year round in two rivers having similar hydromorphological conditions, whereas in the river affected by an upstream lake waterflow, more species were detected sporadically (42%). For all the rivers, in more than 98% of the occasional detections, the reads abundance represented <0.4% of the total reads per site and per sampling session. Even if the majority of the fish communities remained similar over the year for each of the three rivers, specific seasonal patterns were observed. We studied if the waterflow or the reproduction period had an effect on the observed dynamics. Waterflow, which influences eDNA downstream transportation, had a global influence in taxonomic richness, while the fishes' reproductive period had only an influence on certain species. Our results may help selecting the best sampling strategy according to research objectives. To study fish communities at local scale, seasons of low waterflow periods are recommended. This particularly helps to restraint effects of external eDNA coming from connections with other aquatic environment (tributaries, lakes, wetlands, sewage effluents, etc.). To obtain a more integrative overview of the fish community living in a river basin, high waterflow or breeding seasons are preferable for enhancing species detection probability, especially for rare species.