Detection sensitivity of aquatic species using environmental DNA (eDNA) generally decreases in turbid water but is poorly characterized. In this study, eDNA detection targeted delta smelt (Hypomesus transpacificus), a critically endangered estuarine fish associated with turbid water. eDNA sampling in the field was first paired with a trawl survey. Species-specific detection using a Taqman qPCR assay showed concordance between the methods, but a weak eDNA signal. Informed by the results of field sampling, an experiment was designed to assess how turbidity and filtration methods influence detection of a rare target. Water from non-turbid (5 NTU) and turbid (50 NTU) estuarine sites was spiked with small volumes (0.5 and 1 mL) of water from a delta smelt tank to generate low eDNA concentrations. Samples were filtered using four filter types: cartridge filters (pore size 0.45 μm) and 47 mm filters (glass fiber, pore size 1.6 μm and polycarbonate, pore sizes 5 and 10 μm). Prefiltration was also tested as an addition to the filtration protocol for turbid water samples. eDNA copy numbers were analyzed using a censored data method for qPCR data. The assay limits and lack of PCR inhibition indicated an optimized assay. Glass fiber filters yielded the highest detection rates and eDNA copies in non-turbid and turbid water. Prefiltration improved detection in turbid water only when used with cartridge and polycarbonate filters. Statistical analysis identified turbidity as a significant effect on detection probability and eDNA copies detected; filter type and an interaction between filter type and prefilter were significant effects on eDNA copies detected, suggesting that particulate-filter interactions can affect detection sensitivity. Pilot experiments and transparent criteria for positive detection could improve eDNA surveys of rare species in turbid environments.