Chum salmon fisheries in Skagit Bay, Possession Sound, Central Puget Sound, and Hood Canal were sampled weekly in 1997 and 1998. Stock composition of the samples was estimated by using genetic stock identification (GSI) analysis. Sampling crews extracted heart, liver, and muscle tissue from individual fish landed in the fisheries. Maximum likelihood estimation computed estimates of baseline chum stocks composed in the samples. Fishery samples received a two-step analysis. Lower-than-normal returns in 1997 resulted in some sampling goals not being achieved. Additional sampling taken in 1998 met with better results. Weekly mean estimates were graphically constructed from the time series of data stretching back to 1991 for the Hood Canal and Central Puget Sound catch areas and back to 1994 for the Possession Sound and Skagit Bay catch areas. Correlation coefficients were calculated from weekly stock composition estimates to compare similarities among stock groups. Negative correlation coefficients were computed between at least two stock groups in each fishery area. Overall, the GSI analysis showed that non-local stocks contributed to the terminal area harvest, at times significantly.