Slow motion gave no clear added wind prediction accuracy in two cyclone basins in 2003-2024

We tested whether slow preceding cyclone motion improved estimates of next-day wind gain beyond current wind, position, month and prior wind gain. We fit fixed models on 1982-2002 and scored 2003-2024 in IBTrACS v04r01. The test used 1,303 North Atlantic fixes from 149 storms and 3,609 eastern Pacific fixes from 300 storms. Adding a below-3-m/s indicator reduced mean squared error by 0.124% and 0.213%. The paired five-season block 97.5% intervals were -0.321% to 0.561% and -0.085% to 0.544%. All planned scoring and model checks also had intervals that included zero. Thus the test gave no clear added accuracy under these models. The sample requires complete later tropical offshore windows. It is a retrospective conditional test, not an operational trial. Earlier full-period evidence informed the design. No ocean or atmospheric fields were used, so the result cannot identify a physical cause. Code, source hashes, selected source data, predictions, block draws, a figure and numerical checks are provided.

tropical cyclone · intensity prediction · translation speed · IBTrACS · chronological test · null result · ocean feedback