Evaluation of the 1994 Predictions of the Run-Timing of Wild Migrant Yearling Chinook in the Snake River Basin.

Since 1988, wild salmon have been PIT-tagged under programs conducted by the National Marine Fisheries Service, Idaho Department of Fish and Game, Oregon Department of Fish and Game and the U.S. Fish and Wildlife Service. The detection of tagged individuals at Lower Granite Dam provides a measure, of the temporal and spatial distribution of the wild populations. PIT Forecaster was developed to take advantage of this historical data to predict the proportion of a particular population which had arrived at the index site in real-time and to forecast elapsed time to some future percentile in a migration. This report evaluates the performance of the Least Squares (LS) method and the Synchronized historical pattern matching (SYNC) method used in the PIT Forecaster and two other possible methods of prediction reminiscent of methods recommended by the Fish Passage Center (FPC). Alternative No. 1 bases predictions on the historical proportion of PIT-Tags recovered in a specific year and Alternative No. 2 uses the historical cumulative distribution of smolt predicted of a previous season. Over the entire 1994 season, the LS method had the best prediction performance for both aggregate and individual streams. However, for the first half of the season, Alternative No. 1 was clearly a better predictor. This performance deteriorated for the last half of the season, and the LS method improved prediction for aggregate streams and for the individual streams. The algorithms used by PIT Forecaster provide a better prediction as the season progresses. For the first half of the season, Alternative No. 1 did very well for both index years selected. For the 1995 season, Alternative No. 1 will be combined with the LS method to create an improved predictor.