The literature on abundance of birds of prey is almost devoid of population estimates with statistical rigor. Therefore, we surveyed bald eagle (Haliaeetus leucocephalus) populations on the Crooked and lower Columbia rivers of Oregon and used the double-survey method to estimate populations and sighting probabilities for different survey methods (aerial, boat, vehicle) and bald eagle ages (adults vs. subadults). Sighting probabilities were consistently <1.00, so direct counts were biased estimates of abundance. There was a greater probability of observing adult (0.625) versus subadult (0.494) eagles on the Crooked River for different survey methods, but we did not detect differences in sighting probabilities for adults (0.707) versus subadults (0.683) on the Columbia River. The probability of observing bald eagles from aircraft (0.636) was lower than that from boats (0.758) on the Columbia River, and sighting probabilities from aircraft (0.354) were lower than those from the ground (0.764) on the Crooked River. Precision of population estimates was generally good: most (68%, n = 30) coefficients of variation were <10%, and only 7% were >20%. The results revealed variable and negative bias (percent relative bias = -9 to -70%) of direct counts and emphasized the importance of estimating populations where some measure of precision and ability to conduct inference tests are available. We recommend use of the double-survey method to estimate abundance of bald eagle populations and other raptors in open habitats.