Prior track turning gave no clear next-day Atlantic wind prediction gain in selected offshore windows in 2003-2024

We tested whether prior track turning improves next-day wind-change estimates after storm speed and state are known. We used NOAA IBTrACS v04r01 and fit each test season from earlier seasons. The selected offshore Atlantic sample has 1,697 test windows from 140 storms in 2003-2024. Adding one turning term increased mean squared error by 0.49%. The conditional 95% interval for error reduction was -1.27% to 0.006%. All six planned sample checks also gave higher point errors. Only 45 test windows had turns of at least 60 degrees. Thus, this simple past-turn measure gave no clear added prediction value in this sample. This result does not test the physical effect of ocean cooling or the value of a forecast of future turns. The short record, rare large turns, revised tracks, future sample selection, and fixed-prediction uncertainty limit inference. The package includes source data, hashes, code, all predictions, block draws, a separate geometry check, and a figure.

tropical cyclone · track turning · ocean feedback · chronological prediction · null result · IBTrACS