While environmental DNA (eDNA) is now being regularly used to detect rare and elusive species, detection in lotic environments comes with a caveat: The species being detected is likely some distance upstream from the point of sampling. Here, we conduct a series of seminatural stream experiments to test the sensitivity of new digital droplet PCR (ddPCR) to detect low concentrations of eDNA in a lotic system, measure the residence time of eDNA compared to a conservative tracer, and we model the transport of eDNA in this system. We found that while ddPCR improves our sensitivity of detection, the residence time and transport of eDNA does not follow the same dynamics as the conservative tracer and necessitates a more stochastic framework for modeling eDNA transport. There was no evidence for differences in the transport of eDNA due to substrate type. The relatively large amount of unexplained variability in eDNA transport reveals the need for uncovering mechanisms and processes by which eDNA is transported downstream leading to species detections, particularly when inferences are to be made in natural systems where eDNA is being used for conservation management.