Monthly fits reproduce published opposite tide-offset trends at two US gauges in retained 1960-2025 records

A separate monthly implementation reproduces a published annual tide-offset result on the same NOAA source data. We used 4,435 accepted station-months at six US gauges in 1960-2025. The tested difference is MHHW plus MLLW minus twice monthly mean sea level. Fits include season, linear time, and a fixed 18.61-year cycle. The difference fell at Charleston by 0.374 mm/year, with a 95% interval of -0.455 to -0.293, and rose at San Francisco by 0.271 mm/year, with an interval of 0.190 to 0.352. Both pass a six-site Holm rule; four other sites do not. The published complete-year estimates were -0.392 and +0.308. Thus a monthly fit with partial-year records and different error estimates retains the two-site conclusion. A 1975 start retains both signs. Split-span fits do not support steady change across all decades, and omission of the nodal term makes both full-record intervals include zero. This is an implementation and inclusion-sensitivity check on shared data, not new independent observations or a new physical mechanism. Missing months, source changes, and model assumptions limit inference. The package contains exact raw bytes, hashes, code, all fits, a figure, and numerical checks.

tide gauges · tide shape · NOAA CO-OPS · lunar nodal cycle · trend uncertainty