Effective monitoring strategies are key for aquatic species conservation, but traditional methods often require significant resources, especially in large aquatic systems. The emergence of quantitative environmental DNA (eDNA) is a promising alternative. Yet, few studies have evaluated the possibility of quantifying Atlantic salmon abundance with eDNA when individuals are at low abundance and occur in large river systems. In this study, we tested the efficacy of eDNA to monitor and quantify daily variation in smolts counts during their downstream migration in a large river system with low abundance of smolts. During the 2021 and 2022 downstream migrations, trap nets were used to conduct a daily census of smolts in the Romaine and Puyjalon rivers (Québec, Canada) while eDNA samples were collected daily over a transect perpendicular to the riverbank. Using quantitative real-time qPCR, we showed that discharge-corrected eDNA concentrations were positively correlated with daily smolt counts for both years. In addition, we found that controlling for temperature and precipitation improved model transferability between years, showing the importance of considering environmental correlates when using eDNA for abundance quantification. Finally, smolt counts were correlated with eDNA concentrations on the same day, but not with eDNA concentrations one or two days prior, highlighting the capacity of the model to track daily fluctuations in smolt abundance. Our results underscore the potential of using eDNA to monitor Atlantic salmon in large river systems with low smolt abundance when the river hydrology and environmental conditions are documented.