Uncertainties in tropical cyclone landfall decay

Tropical cyclones cause major losses when they make landfall, and several studies suggest their destructive potential has increased and may continue to rise in a warming climate. Recent high-profile studies reported a significant slowdown in the decay of landfalling tropical cyclones (e.g., North Atlantic hurricanes; western North Pacific systems), attributing the effect variously to thermodynamic (sea surface temperature) or dynamic changes. However, other work questioned the methodologies and robustness of those findings, suggesting decay depends strongly on landfalling track modes and effective moisture supply from the ocean. The central research question here is whether a slowdown in post-landfall decay is evident and globally consistent. The purpose is to provide a systematic global climatology of landfall decay and to assess how inferences depend on basin, period, dataset, event selection criteria, land–sea mask resolution, and statistical technique. This analysis is important for accurately diagnosing climate signals, guiding risk assessment, and informing adaptation strategies.

The introduction surveys key prior studies: Li & Chakraborty (2020) reported a significant slowdown of decay for North Atlantic landfalling hurricanes (1967–2018) linked primarily to rising sea surface temperatures. Song et al. (2021) reported a slowdown for western North Pacific landfalling cyclones on the Asian continent (1966–2018), attributing it more to dynamic than thermodynamic changes. Chan et al. (2022) challenged the methodology and robustness of Li & Chakraborty, noting calculation inadequacies and emphasizing the role of landfalling track modes and effective oceanic moisture supply. Phillipson & Toumi (2021) did not find significant reductions in decay constants, contradicting some prior claims. Broader literature highlights increasing destructive potential of landfalling TCs and track migrations toward coasts, as well as known discrepancies among best-track datasets and improvements since the satellite era. This mixed evidence motivates a comprehensive reassessment.

Data: The study uses authoritative best-track datasets: HURDAT2 (North Atlantic, Eastern North Pacific), and CMA and JTWC tracks from IBTRACS (western North Pacific). For South Indian Ocean and South Pacific, JTWC is used to maintain homogeneity. JMA and HKO were excluded for WNP due to fewer identified landfalls. Reanalysis and ancillary datasets include ERA5 (0.25°, 1980–2020) for winds and subtropical highs, and ETOPO1 topography/bathymetry to construct a land–sea mask.

Land–sea mask: ETOPO1 was regridded to 0.1° using bilinear interpolation. For each best-track position, the nearest 9 grid points were examined; if ≥3 points had elevation ≥1 m, the TC position was considered over land. This approach better captures near-coast landfalls and filters small islands/inland water, avoiding contamination seen with coarser 1.25° masks used in some prior studies.

Landfall event definition: Three criteria were applied: (1) two landfall intensity thresholds were analyzed separately: ≥34 kt (gale force) and ≥64 kt (hurricane force), the latter following prior work to downweight weak, poorly observed inland events especially before the 1980s; (2) at least four continuous synoptic inland points (0000, 0600, 1200, 1800 UTC), ensuring ≥18 h over land; (3) exclude events that intensified, had no change, or underwent extratropical transition within the first 18 h past landfall.

Study periods: Trends were evaluated over three periods for each basin/dataset: from the start of the best-track record; from 1967 onward (to match prior studies); and from 1980 onward (global satellite era, generally more reliable).

Decay timescale τ: Assuming exponential post-landfall wind decay V(t) = V(0) e^(−τ t), τ was estimated via regression using the first four continuous synoptic inland points. Larger τ indicates slower decay.

Effective oceanic area of moisture supply (A_o): Defined as the average oceanic area within a 200-km radius of the TC center from the second to fourth synoptic times past landfall.

Landfalling track modes: Following Chan et al. (2022), events were categorized into “hard-strike” and non–“hard-strike” modes, with the latter subdivided into cross-over, lingering, catwalk, and skirting, based on post-landfall track geometry and air–sea interaction potential. Over 98% of events were categorized.

Statistical approaches: Trends were computed for both season-level (seasonal mean τ, equally weighting seasons) and event-level τ (weighting by landfall frequency). Linear regression was used, with reported slopes, p-values, and sample sizes. Sensitivity to basin, dataset, period, intensity threshold, land–sea mask resolution, and statistical technique was assessed.

• Basin dependence: Trends differ by basin. For example, during 1967–2020 in the North Atlantic (HURDAT2), season-level τ for landfalls ≥64 kt shows a significant increase of 0.42 h/yr (p<0.05; n=39 seasons), whereas other basins show mostly insignificant or decreasing trends over similar periods. The reported slowdowns are not globally ubiquitous.
• Data dependence (agency differences): Using different best-track sources yields different trends in the same basin and period. In the WNP for ≥34 kt landfalls: JTWC (1945–2019) shows a significant decreasing trend of −0.3 h/yr (p<0.01; n=70), whereas CMA (1950–2019) shows an insignificant trend of −0.03 h/yr (p=0.75; n=70). For ≥64 kt landfalls (1980–2019): CMA shows −0.6 h/yr (p<0.01; n=30) while JTWC shows −0.35 h/yr (p=0.15; n=31). Such inconsistencies likely stem from heterogeneous operational procedures and intensity estimation discrepancies.
• Study period dependence: Trends and their significance vary with the chosen start date, and can even reverse sign. For example, in the North Atlantic, a decreasing trend over 1851–2020 becomes increasing over 1967/1980–2020. Similar reversals appear in the WNP depending on period. This suggests substantial decadal to multidecadal variability.
• Landfall intensity dependence: Trend inferences depend on the landfall intensity threshold. In the North Atlantic (1967–2020), ≥64 kt shows a significant increase (0.42 h/yr, p<0.05), whereas ≥34 kt is insignificant (0.15 h/yr, p=0.49). In the WNP (1967/1980–2019), ≥34 kt trends are increasing while ≥64 kt trends are decreasing.
• Land–sea mask dependence: Using a finer 0.1° land–sea mask alters event selection and trends relative to a coarse 1.25° mask (as in SK), which can erroneously include over-sea points and change trend assessments, supporting prior critiques.
• Statistical technique dependence: Season-level vs event-level analyses can disagree. For NA ≥64 kt (1967–2020), the season-level increasing trend (0.42 h/yr, p<0.05) becomes insignificant at event-level (0.17 h/yr, p=0.47). Conversely, WNP ≥34 kt (CMA, 1967/1980–2019) shows insignificant increases at season level (0.13–0.16 h/yr, p>0.05) but significant increases at event level (0.25–0.36 h/yr, p<0.05).
• Effective oceanic area of moisture supply (A_o) as a key control: Globally, τ correlates positively with A_o. For ≥34 kt events, r=0.35 (N=1287) and r=0.36 with extreme τ>200 h excluded (N=1284). For ≥64 kt events, r=0.44 (N=549) and r=0.55 with τ>200 h excluded (N=546). Larger A_o (typical of non–hard-strike modes) is associated with slower decay due to greater sea-surface enthalpy flux and reduced land-surface friction.
• Weak vs strong events: The A_o–τ correlation strengthens when excluding weaker landfalls (≥64 kt subset), implying greater uncertainty and variability among weak (34–64 kt) landfalls. In the WNP (CMA), weak “hard-strike” events on China/Indochina show significantly increasing τ over 1950–2019/1967–2019/1980–2019 (e.g., season-level trends ~0.31–0.58 h/yr; p<10⁻³), while strong “hard-strike” events do not show similar increases.
• Large-scale steering influence: Changes in subtropical highs since 1980 are consistent with shifts in landfalling track modes and trends. Over the NA/ENP, an eastward-retreating subtropical high favored more non–hard-strike landfalls (longer τ), while a westward-extending WNP subtropical high favored more hard-strike landfalls on China (shorter τ).

The study addresses whether landfall decay slowdown is a robust, global signal attributable to warming. Findings show that claims of a universal slowdown are not supported: trends vary by basin, dataset, period, intensity threshold, land–sea mask resolution, and analysis method. This undermines the hypothesis that rising SST alone universally lengthens inland decay. Instead, a physically grounded control emerges: the effective oceanic area of moisture supply (A_o), closely linked to landfalling track modes and large-scale steering (e.g., subtropical highs), exerts a primary influence on decay rates. Regions and periods with more non–hard-strike tracks (larger A_o) exhibit longer decay timescales, while hard-strike landfalls (small A_o) decay faster. The diversity and reversals of trends, together with small-sample basins, suggest that decadal to multidecadal variability can mask or mimic long-term trends. Thus, any climate trend inference requires comprehensive, basin-aware, and method-consistent analyses rather than isolated snapshots.

This work provides a systematic global climatology of tropical cyclone post-landfall decay and clarifies that reported slowdowns are neither universal nor robust across datasets, basins, study periods, or methodologies. A key contribution is identifying the effective oceanic area of moisture supply (A_o), governed by landfalling track modes and large-scale steering, as a generic control on decay timescales, with strong global correlations. The study cautions against asserting a universal climate-driven slowdown at present, while not excluding possible long-term trends. Future work should: (1) expand and homogenize best-track datasets with longer records and improved cross-agency consistency in intensity estimates; (2) explicitly account for decadal–multidecadal variability and seek periodicities rather than only linear trends; (3) use multi–high-resolution climate models to assess projected changes in track modes, A_o, and decay behavior; and (4) refine land–sea masks and event definitions to ensure robust inland decay diagnostics.

• Small sample sizes in certain basins (Eastern North Pacific, North Indian Ocean, South Pacific) reduce statistical power and trend robustness.
• Heterogeneities among best-track datasets (e.g., CMA vs JTWC intensity estimates) lead to data-dependent and sometimes contradictory trends.
• Trend inferences are sensitive to study period selection, given pronounced decadal–multidecadal variability.
• Event selection criteria and land–sea mask resolution materially affect which cases are included, influencing trend estimates.
• Weak landfalling TCs (34–64 kt) have greater structural uncertainty and observational limitations, especially pre-satellite era, contributing to higher variance in τ.
• Despite the comprehensive analysis, the study cannot rule out long-term climate trends; current variability and data issues may obscure subtle signals.