In-season methods that produce accurate and timely forecasts of returning salmon abundances allow fisheries managers to alter fishing plans in order to meet conservation and harvest objectives. In-season methods are challenged by variability in catch statistics due to factors external to abundance, specifically, fluctuations in the migration timing of target and co-migrating stocks. I apply genetic stock identification (GSI) data to develop catch indices for the five Fraser Chinook management units, and use these indices to forecast returns for each management unit according to four in-season model forms. I evaluate models using three performance measures to determine forecasting errors. Results show that forecasts for Spring 52 and Summer 52 Chinook can be produced with reasonable accuracy early in the fishing season. Forecasts of Spring 42, Summer 41, and Fall Chinook are less accurate. Results indicate that this technique shows promise for providing accurate and timely forecasts for the five Fraser Chinook management units, particularly as additional years of data are GSI-analyzed.