Prior pressure change did not lower next-day wind test error in two offshore cyclone samples in 2003-2023

We tested whether prior 24-hour pressure change adds next-day wind information beyond current pressure, wind, prior wind change, position and month. We used fixed IBTrACS v04r01 records and fit models only in 1982-2002. The Atlantic test had 1,655 windows from 133 storms in 2003-2023. The mean squared error gain from pressure change was -0.95%, with a conditional 95% three-year block interval of -1.78% to -0.05%. In the eastern Pacific test, the gain was -0.59%, with an interval of -1.69% to 0.74%. These results test a simple retrospective method. They do not establish a pressure-driven process or an operational forecast gain. Future eligibility, revised and related wind-pressure estimates, a fixed fit, and 21 test years limit inference. The package includes source columns, hashes, code, predictions, sensitivity checks, a figure, and a separate numerical audit.

tropical cyclone · pressure change · wind prediction · IBTrACS · chronological test · offshore