Adding bottom temperature reduced estimation error on 158 fish surveys at Maizuru Bay in 2012–2023

Can seasonal phase improve estimates of fish species shares at similar water temperatures? We used open visual survey data from Maizuru Bay, Japan. A fixed library had 240 surveys from 2002–2011. The later test had 288 surveys from 2012–2023. We kept 158 targets with at least five earlier matches from each seasonal phase within 2 degrees C at both depths. Each estimate used five earlier square-root species-share vectors. In the initial test, adding phase to both-temperature matching increased squared Hellinger loss by 3.20%. An added exploratory comparison kept the same 158 targets. Adding bottom temperature to surface-temperature matching reduced loss from 0.335310 to 0.308471, or 8.00%. The nominal 95% three-year-block range was 5.01% to 10.92%. The gain stayed positive with three or ten matches, equal species shares, and jack mackerel removed. It was also positive under 199 random tie rules; its median was 5.53%. A simple calendar mean had similar error. The phase result changed with species weights. These are concurrent composition estimates at one site, not advance forecasts or causal temperature effects. Ranges condition on the fixed earlier library and selected targets. The package preserves source tables and notices, all match records, predictions, results, figures, code, and a successful clean repeat.

fish community · water temperature · Maizuru Bay · seasonal phase · composition · Hellinger distance · reproducibility · chronological test