Shad passage at Columbia River dams stopped advancing after about 1988 while river temperature increased, and 1950-1993 temperature slopes did not hold in 1994-2025

Quinn and Adams (1996) reported that American shad passage at Bonneville Dam moved much earlier in 1938-1993 and followed spring river temperature, while sockeye salmon moved earlier less and passed in warmer water. We tested these results out of sample with DART daily counts and dam temperatures for 1939-2025 at Bonneville, The Dalles, McNary and Rock Island dams. We fitted models on 1950-1993 and tested them on 1994-2025, as written in an analysis plan before the test. The shad 50% passage date at Bonneville moved earlier by 9.6 days per decade in 1950-1993 (95% CI 8.1 to 11.1) but did not move earlier in 1994-2025 (+1.5, 95% CI -0.3 to +3.4). McNary gave the same result, and an exploratory two-segment fit put the change near 1988 at three dams. River temperature before passage continued to increase (+0.30 C per decade in 1994-2025). The shad temperature slope changed from -9.5 days per C in 1950-1993 to -1.1 days per C in 1994-2025 (difference +8.4, 95% CI 5.7 to 11.1). Thus the temperature at shad passage changed from a decrease of 0.45 C per decade to an increase of 0.45 C per decade. Sockeye timing also stopped moving earlier at Bonneville and McNary, but continued at Rock Island. Higher discharge gave later passage in both periods for both species. The 1950-1993 temperature model gave a small skill gain over the mean for 1994-2025 shad dates (0.18, 95% CI 0.03 to 0.32) but predicted dates 9 days too late; this level error depends on a 0.04-0.77 C offset between the old DART record and a USGS sensor. Limitations: one river system, upstream dams count many of the same fish, and the end of the shad advance occurred at the same time as the end of a more than 100-fold shad population increase, so the cause is not identified.

American shad · sockeye salmon · migration timing · phenology · river temperature · Columbia River · Bonneville Dam · out-of-sample test · climate change · DART fish counts · observation bias · replication