1. Quantifying habitat associations and threats to the persistence of imperilled species is a fundamental step for initiating species recovery efforts, but the traits associated with species imperillment (e.g. rarity and vulnerability to stressors) also limit the ability to empirically inform recovery strategies. Novel sampling designs and modelling approaches are therefore needed to quantitatively assess habitat associations and the threats to species persistence. 2. To improve the understanding of habitat associations and threats for one of the rarest endangered freshwater fish species in Canada, northern madtom (Noturus stigmosus), two-species occupancy models were developed with two invasive gobies: round goby (Neogobius melanostomus) and tubenose goby (Proterorhinus marmoratus). Models were parameterized with data from a multi-year benthic trawling survey of two large Great Lakes tributaries in southern Ontario, Canada, and subsequently used to evaluate the need for reducing measurement error with future sampling efforts. 3. The probability of detecting northern madtom in the St. Clair (0.163) and Thames (0.194) rivers was low compared with round goby (St. Clair, 0.827; Thames, 0.833) and tubenose goby (St. Clair, 0.297). The best occupancy models indicated a negative association between northern madtom and round goby in the St. Clair River and the importance of gravel substrate for northern madtom. 4. Up to 16 repeated non-detections using benthic trawls are needed to be 95% confident that northern madtom is absent at a site, indicating that current sampling approaches are likely inadequate. 5. Despite low detection probabilities, intensive trawling surveys combined with the two-species occupancy modelling framework provided vital information for describing habitat associations for northern madtom and identified a significant negative association with round goby. Nonetheless, alternative sampling methods Reproduced with the permission of the Minister of Fisheries and Oceans Canada