We determined the habitat usage and habitat connectivity of juvenile Chinook ( Oncorhynchus tshawytscha) and coho ( Oncorhynchus kisutch) salmon in continental shelf waters off Washington and Oregon, based on samples collected every June for 9 yr (1998–2006). Habitat usage and connectivity were evaluated using SeaWiFS satellite-derived chlorophyll a data and water depth. Logistic regression models were developed for both species, and habitats were first classified using a threshold value estimated from a receiver operating characteristic curve. A Bernoulli random process using catch probabilities from observed data, i.e. the frequency of occurrence of a fish divided by the number of times a station was surveyed, was applied to reclassify stations. Zero-catch probabilities of yearling Chinook and yearling coho salmon decreased with increases in chlorophyll a concentration, and with decreases in water depth. From 1998 to 2006, ∼ 47% of stations surveyed were classified as unfavorable habitat for yearling Chinook salmon and ∼ 53% for yearling coho salmon. Potentially favorable habitat varied among years and ranged from 9 856 to 15 120 km2 (Chinook) and from 14 800 to 16 736 km2 (coho). For both species, the smallest habitat area occurred in 1998, an El Niño year. Favorable habitats for yearling Chinook salmon were more isolated in 1998 and 2005 than in other years. Both species had larger and more continuous favorable habitat areas along the Washington coast than along the Oregon coast. The favorable habitats were also larger and more continuous nearshore than offshore for both species. Further investigations on large-scale transport, mesoscale physical features, and prey and predator availability in the study area are necessary to explain the spatial arrangement of juvenile salmon habitats in continental shelf waters. [ABSTRACT FROM AUTHOR]