Fifteen years of video census data from a fishway at Winchester Dam on the North Umpqua River, Oregon, were used to evaluate sampling designs to estimate abundance of spring‐ and fall‐run Chinook Salmon Oncorhynchus tshawytscha, summer‐ and winter‐run steelhead O. mykiss, Coho Salmon O. kisutch, and Pacific Lamprey Entosphenus tridentatus. Five probabilistic sampling designs were evaluated via simulation, with variation in the number of days and the number of hours within a day that were sampled over a 1‐year period. For most species, stratified one‐stage and two‐stage cluster designs were more accurate at estimating abundance than simple random sampling designs. There was very little gain in accuracy beyond sampling 8 h per day for all species. The stratified two‐stage cluster nonuniform probability design was more accurate than the stratified two‐stage cluster uniform probability design at estimating the abundance of steelhead, spring Chinook Salmon, and Coho Salmon, whereas using uniform probabilities resulted in more accurate estimates of abundance for Pacific Lamprey and fall Chinook Salmon. Additionally, the stratified and nonuniform probability designs can be adjusted for high‐priority species through allocation of the sample and assigning selection probabilities of secondary sampling units that optimize efficiency for those species. The consistency in patterns observed among species suggests that the results of this study can be applied to other systems where the abundance of multiple species is of interest. Received December 5, 2014; accepted March 28, 2015