Published 1977-2001 temperature slopes did not forecast 2002-2025 mean latitudes of North Sea demersal fishes better than no change in two ICES trawl surveys

Perry et al. (2005) reported that the mean latitude of North Sea demersal fishes moved with winter bottom temperature in 1977-2001. We used their published slopes for 14 species as forecasts for 2002-2025, when the 5-year mean winter sea surface temperature was 0.52 C higher. We tested the forecasts with ICES DATRAS data from two surveys: the international bottom trawl survey in quarter 1 (1980-2025, 13,462 hauls) and in quarter 3 (1991-2025, 8,945 hauls; English hauls only as the original source survey). To control observation bias, we used only standard GOV day hauls with all species recorded, averaged catches per ICES rectangle, and kept a fixed set of rectangles. The published slopes did not forecast better than a no-change forecast. The mean log ratio of squared forecast error was -0.00 (95% block-bootstrap interval -0.04 to +0.10) in quarter 1 and +0.30 (+0.11 to +0.48) in quarter 3. In 22 analysis variants, no interval was fully below zero and 13 were fully above zero. In quarter 1, 11 of 13 species moved in the predicted direction, but this result did not stay with log-catch or occupancy weights and did not occur in quarter 3 (7 of 14). Predicted and observed shift sizes were not correlated. Slopes fitted again to 1980-2001 data and time-trend extrapolation also did not beat no change. Limitations: the surveys differ from the original survey in season or period, SST is a proxy for bottom temperature, fishing effects are not separated, and catch-weighted latitude is sensitive to large catches. The result shows that the published species slopes did not transfer to the next 24 years; it does not show that warming has no effect.

North Sea · demersal fish · distribution shift · climate change · out-of-sample forecast · replication · ICES DATRAS · IBTS · sea temperature · observation bias · null result