Forward-Only Temporal-Coherence Occupancy Regimes in Atlantic Tropical Cyclone Intensity Evolution (1851–2024)

Nathan Howell 

We test whether a forward-only state derived solely from maximum sustained wind speed (VMAX) provides a reproducible description of tropical cyclone intensity persistence and historical rapid-intensification (RI) stratification. Atlantic HURDAT2 records for 1,991 storms during 1851–2024 were evaluated with causal 3-, 5-, and 9-advisory tendencies. A state was confirmed only when the current and previous advisory shared the same nonzero common direction; reversals and missing required lags interrupted confirmation, and the first coherent advisory received no retroactive occupancy. The V2 analysis dataset contains 54,174 advisories, of which 15,542 are confirmed active, totaling 93,252 h. Event-ever RI, defined as a VMAX increase of at least 30 kt over an exact 24-h interval, occurred in 525 storms. Among 524 storms with at least 72 h of occupancy, 277 underwent RI (52.86%), compared with 26.37% across the basin archive. This conditional difference is descriptive, not forecast skill. In a shared-sample duration-controlled comparison, simple five-advisory persistence had stronger tested model fit than the three-window state (AIC 1896.01 versus 1991.56), so broad superiority of the multihorizon construction was not established. For 496 storms with both times observed, RI onset minus confirmed activation had a median of −24 h (interquartile range −60 to 0 h); confirmed activation was therefore not generally an advance RI marker. Incremental association beyond the tested environmental covariates was also not established. The construct is a reproducible VMAX-only temporal baseline for testing whether independent physical observations add information beyond persistence and lifecycle history.