Environmental DNA (eDNA) offers a sensitive, non-invasive alternative to conventional fish sampling, but its effectiveness and cost have not been appraised for the fishes of south-western Australia. We evaluated active and passive eDNA collection alongside fyke netting in the Djarlgaroo Beeliar (Canning River), an urban river in Perth, south-western Australia. Using hierarchical models that accounted for imperfect detection, we compared the ability of these methods to describe the native and alien fish assemblage including a newly invading alien species (Pearl cichlid). Active eDNA metabarcoding detected 11 species, including one not captured by fyke netting, and produced higher and more consistent detection probabilities across species than either fyke nets or passive eDNA – but its performance was suboptimal for the bluespot goby (native) and Common carp (alien). Active eDNA was particularly superior when monitoring the Pearl cichlid, being almost three times more effective at detection per replicate than fyke netting. This improved performance offset the costs associated with laboratory analysis making active eDNA the most cost-effective method to monitor this low abundance species. Passive eDNA performed similarly to fyke netting and was considerably less effective than active filtration; however, its performance may have been constrained by a short deployment period (4hrs). We also found that capture probability declined with pool volume for fyke nets and passive eDNA. eDNA capture and persistence are likely to vary with flow, water quality, habitat dimensions and sampler design and further field trials across contrasting environments are needed before passive sampling can realise its potential as a simple, low-cost monitoring approach in the region. Our findings demonstrate that eDNA can substantially improve monitoring, but programs must match the method and sampling effort to the species, environment and management objective. This research was funded by the Australian Government’s National Environmental Science Program.