Linear annual-cycle drift raised next-year monthly sea-level error by 4% at 217 gauges in 2010-2024

We tested whether linear change in the annual sea-level cycle improves next-year monthly mean sea-level forecasts. We used a fixed PSMSL RLR archive and 217 gauges with at least 95% accepted months in both 1980-2009 and 2010-2024. Each forecast used the previous 20 years. The baseline had a mean trend and calendar-month effects. Adding time times annual sine and cosine terms raised test mean squared error by 4.01%. The conditional 95% three-year block interval for the increase was 1.12% to 7.15%. Root mean squared error rose from 92.76 to 94.61 mm. Only five gauges had lower error. Equal weighting of 50 occupied 10-degree cells also gave higher error. Ten-year and 30-year training checks gave the same sign. A shape diagnostic that removes the future annual mean also gave higher error. Thus this simple change model did not improve these forecasts. The result does not show no seasonal change or identify its cause. The short test, conditional uncertainty, final data revisions, future availability selection, and uneven gauge locations limit inference. Code, input rows, source hashes, forecasts, an audit, and a figure are included.

sea level · seasonal cycle · tide gauges · forecast evaluation · PSMSL · annual harmonic