[Paper Review] Genuine Increases in Tropical Cyclone Intensities

Ivo Welch|arXiv (Cornell University)|Mar 8, 2026

Tropical and Extratropical Cyclones Research0 citations

TL;DR

Using extended ADT-HURSAT v7 data up to 2023, the study shows that the rise in tropical cyclone intensities is now driven by increases in strong storms (C3–C4), not solely by a decline in weak storms (C1).

ABSTRACT

Kossin et al. (2020) report a rising ratio of satellite observations of major C3-C5 storms relative to all C1-C5 storms from 1979 to 2017. Decomposing their R = N(C3+)/N(C1+) statistic into per-category shares shows that their trend was driven primarily by fewer C1 rather than more C3-C5 observations. From the first half to the second half of their sample period, their per-year C1 observations fell by 17%. However, extending the record through 2023 greatly changes the picture. Although the relative decline in C1 observations persists, C3 and C4 observations now increase, too. The signal about the intensification of storms now becomes genuine in the extended sample, in that it is driven no longer only by fewer weak but now also by more strong tropical cyclone observations.

Motivation & Objective

• Assess whether the Kossin ratio R = NC3+/NC1+ reflects true intensification or is driven by changes in weaker-storm observations.
• Update the Kossin analysis with ADT-HURSAT v7b/v7 data extending to 2023 to test the robustness of earlier conclusions.
• Decompose R into per-category shares to identify whether increases come from stronger storms, weaker storms, or both.
• Evaluate regional basin differences in the intensity trend and discuss potential methodological or measurement artifacts.

Proposed method

• Use NCEI ADT-HURSAT v7b dataset (1978–2025) with ADT v9.0 to obtain continuous 1-kt wind observations and Saffir-Simpson thresholds.
• Compute R = NC3+/NC1+ and its decomposition into category shares NCk/NC1+ (k = 1…5).
• Decompose ∆R into −∆(NC1/NC1+) −∆(NC2/NC1+) to attribute changes to category shifts.
• Compare early (1979–1997/1998) vs late (1998–2017) and extended (1979–2023) periods with per-year rates and triad smoothing.
• Apply Theil-Sen slopes and Mann-Kendall tests to assess trends in category shares and R.
• Conduct per-basin decomposition to identify regional patterns.

Experimental results

Research questions

• RQ1 Does the extended record (through 2023) show genuine increases in high-intensity tropical cyclones (C3–C4) beyond what can be explained by declines in weak storms (C1)?
• RQ2 How do per-category shares (C1–C5) contribute to changes in the R statistic over time?
• RQ3 Are the intensity trends consistent across basins, or are there regional differences that affect the global signal?
• RQ4 To what extent might measurement artifacts or ADT version changes influence observed trends?

Key findings

• In the extended 1979–2023 record, the C3 and C4 observations rise, while C1 declines persist but are smaller, leading to a genuine intensification signal.
• The R statistic increases to ∆R ≈ 0.046 (vs ∆R ≈ 0.0267 in the earlier Kossin period), indicating stronger support for intensity increases in the extended data.
• C1 share declines at about 2.5 percentage points per decade; C3 and C4 shares each rise by about 1 percentage point per decade; C5 shows a non-significant decline.
• Regional results show significant trends in the North Atlantic and Southern Indian basins, while the Western Pacific shows essentially no trend.
• Overall, the mechanism shifts from “fewer weak storms” driving the ratio to a combination of “fewer weak storms and more strong storms” in the extended period.