Start-year urchin counts gave an uncertain mean gain in 1,759 next-year kelp forecasts at 200 California survey sites

Sea urchins can remove kelp, but an association measured in the same survey does not establish forecast skill. We tested whether start-year adult urchin density improves one-year kelp-state forecasts beyond start-year kelp density, kelp state and site history. We reused public KFM, PISCO and SBC LTER data. For each program and target year, models used only earlier target years for training. The main test has 1,759 forecasts at 200 revisited California survey sites, with program-specific target years from 1993 to 2025. A state has kelp density of at least 0.1 plants per square metre. The equal-program mean Brier score fell from 0.09882 to 0.09531 after adding start-year urchins. The relative gain was 3.55%, with a 95% interval of -2.60% to +9.08% from crossed site and three-year calendar-block draws. Thus, the tested models did not show a clear mean gain. The older KFM data supplied 76% of the mean gain; without them, the point estimate was 1.28%. Target-year urchins gave a larger point estimate, but those counts would be unavailable when a forecast is issued. Planned model and state checks gave start-year gains of 0.81% to 4.96%. The package includes exact source data, all forecasts, uncertainty draws, figures and offline code. Independent calculations reproduced all primary probabilities within 1.2e-8. This is a conditional forecast benchmark at sampled sites. It does not estimate an effect of urchin removal.

kelp · sea urchins · ecological forecasting · chronological validation · reef surveys · Brier score · open data