Genuine Increases in Tropical Cyclone Intensities

While Kossin et al. (2020) attributed a rising ratio of major tropical cyclones to a decline in weaker Category 1 storms, extending the data through 2023 reveals a genuine intensification trend driven by both fewer weak storms and an actual increase in strong Category 3 and 4 storms.

The Gist

Imagine you are looking at a giant bucket of marbles. Some marbles are small and weak (Category 1 storms), some are medium, and some are huge, powerful boulders (Category 3, 4, and 5 storms).

For a long time, scientists have been watching this bucket to see if climate change is making the storms inside it stronger. A famous study by a team led by Kossin looked at data from 1979 to 2017 and concluded: "Hey, the bucket has way more big boulders now!"

They calculated this by looking at a specific ratio: How many big boulders are there compared to the total number of marbles?

The Old Story (1979–2017): The "Missing Small Marbles" Trick

This new paper by Ivo Welch says, "Wait a minute. Let's look closer at why that ratio went up."

When you look at the 1979–2017 data, the ratio of big storms went up. But Welch found that it wasn't because the ocean suddenly started spawning more giant hurricanes. It was mostly because the number of small, weak hurricanes dropped significantly.

The Analogy:
Imagine a classroom.

• 1979: 100 students. 50 are quiet kids (weak storms), 50 are loud kids (strong storms). The "Loud Ratio" is 50%.
• 2017: 80 students. Only 20 are quiet kids, but 60 are loud kids. The "Loud Ratio" is 75%.

The ratio of loud kids went up! But did the loud kids get louder? Not necessarily. The ratio went up mostly because 20 quiet kids stopped showing up to class. The scientists thought the "loudness" was increasing, but it was actually just a disappearance of the "quiet" ones.

This could be a real weather change, or it could be a glitch in the satellite cameras (like a camera getting better at seeing big storms but missing the small, fuzzy ones).

The New Story (1979–2023): The "Real Growth"

Welch updated the data to include the years 2018 through 2023. He used a newer, sharper version of the satellite data (like upgrading from a standard definition TV to 4K).

When he added these recent years, the picture changed completely.

1. The quiet kids are still disappearing: The number of weak storms (Category 1) is still dropping.
2. BUT, now the loud kids are actually multiplying: The number of Category 3 and 4 storms is increasing.

The New Analogy:
Now, the classroom has 80 students total (still fewer than before).

• There are only 15 quiet kids (down from 20).
• BUT, there are now 65 loud kids (up from 60).

Now, the ratio of loud kids is high for two reasons:

1. Fewer quiet kids are there.
2. More loud kids are actually showing up.

The Bottom Line

The paper argues that for the first part of the record (1979–2017), the evidence for stronger storms was shaky because it relied on the absence of weak storms.

But with the new data through 2023, the evidence is genuine. We are now seeing a "double whammy":

• We are seeing fewer weak storms.
• AND we are seeing a genuine, observable increase in the number of super-strong, dangerous storms.

Why does this matter?

It changes how we understand the problem.

• Old View: "Maybe our satellites just got better at ignoring small storms, so it looks like storms are getting stronger."
• New View: "No, the small storms are vanishing, and the big, destructive monsters are actually becoming more frequent."

This suggests that the ocean is truly shifting its energy, pushing more storms into the "danger zone" (Categories 3, 4, and 5), which is bad news for coastal communities. The trend is no longer just an optical illusion; it's a real intensification.

Technical Summary

Here is a detailed technical summary of the paper "Genuine Increases in Tropical Cyclone Intensities" by Ivo Welch (UCLA, March 2026).

1. Problem Statement

The paper addresses a debate regarding the interpretation of tropical cyclone (TC) intensification trends. Kossin et al. (2020) reported a rising ratio ($R$) of major tropical cyclones (Categories 3–5) relative to all tropical cyclones (Categories 1–5) from 1979 to 2017. They interpreted this as evidence of a shift in the intensity distribution toward stronger storms due to increasing environmental potential intensity.

However, Welch argues that the ratio statistic $R = N_{C3+} / N_{C1+}$ is susceptible to "denominator effects." An increase in $R$ can be driven by:

1. Numerator effect: An actual increase in the number of strong storms ($C3+$).
2. Denominator effect: A decrease in the number of weak storms ($C1$), which shrinks the denominator even if the numerator remains static.

Welch posits that Kossin et al.'s original finding (1979–2017) was driven primarily by a decline in Category 1 observations rather than a genuine increase in major storms. The paper aims to test whether this trend holds when the dataset is extended through 2023 using updated data.

2. Methodology

The study utilizes the ADT-HURSAT v7b dataset, an updated version of the dataset used by Kossin et al. (which used v6).

• Data Source: NOAA/NCEI's ADT-HURSAT v7b, covering 1978–2025 with 4,852 storms across six basins.
  • Improvements over v6: Uses Advanced Dvorak Technique (ADT) version 9.0 on HURSAT v07b imagery. Wind speeds are continuous (1-knot resolution) rather than binned in 5-knot intervals. This results in systematically higher wind speed estimates (68% more observations exceed the cyclone threshold) but preserves high correlation with v6 ($r > 0.9$).
• Statistical Decomposition:
  • The ratio $R$ is decomposed into additive category shares:
    $$R = \frac{N_{C3}}{N_{C1+}} + \frac{N_{C4}}{N_{C1+}} + \frac{N_{C5}}{N_{C1+}}$$
  • The change in $R$ ($\Delta R$) is exactly equal to the negative sum of the changes in the shares of C1 and C2:
    $$\Delta R = -\Delta\left(\frac{N_{C1}}{N_{C1+}}\right) - \Delta\left(\frac{N_{C2}}{N_{C1+}}\right)$$
  • This allows the author to isolate whether the trend is driven by fewer weak storms or more strong storms.
• Trend Analysis:
  • Time Periods: Compares the original Kossin period (1979–2017) against an extended period (1979–2023).
  • Tests: Uses Theil-Sen estimators for slope calculation and Mann-Kendall tests for significance.
  • Visualization: Kernel density estimates of wind-speed distributions (1-kt resolution) to visualize shifts in the intensity spectrum.

3. Key Contributions

1. Data Extension: Extends the analysis from 2017 to 2023, incorporating the most recent six years of satellite data.
2. Methodological Clarification: Demonstrates mathematically that the Kossin ratio trend can be fully decomposed into changes in specific category shares, separating "denominator effects" (loss of weak storms) from "numerator effects" (gain of strong storms).
3. Dataset Upgrade: Validates that the shift from ADT v6 to v7 does not alter the fundamental trend direction but provides higher resolution and more consistent observations near classification boundaries.

4. Key Results

A. The Original Period (1979–2017)

• Observation: The ratio $R$ increased, but this was primarily driven by a decline in Category 1 (C1) observations.
• Data: Per-year C1 observations fell by 17% (from 341 to 282). Total C1+ observations fell by 9%.
• Interpretation: The intensification signal was largely an artifact of fewer weak storms being detected or reaching cyclone strength, rather than a massive surge in major storms. The surplus in C3–C4 was small compared to the deficit in C1.

B. The Extended Period (1979–2023)

• Observation: The trend shifts significantly. While the decline in C1 observations persists (though slightly less steep at -15%), C3 and C4 observations now show genuine increases.
• Data:
  • Total C1+ observations are nearly flat (-1%).
  • C3 per-year rates increased by 9% (192 to 208).
  • C4 per-year rates increased by 17% (127 to 149).
  • The change in the ratio $\Delta R$ increased from 0.026 (Kossin period) to 0.046 (extended period).
• Visual Evidence: Figure 1 shows that in the extended sample, the "surplus" of observations in the C3–C4 range (red area) roughly matches the "deficit" in the C1 range (blue area). This indicates a genuine shift in the distribution, not just a loss of weak storms.

C. Statistical Significance

• Global Trend: The Mann-Kendall P-value for the global $R$ trend improves from 0.02 (1979–2017) to 0.003 (1979–2023) using 3-year smoothing, and 0.010 using annual data.
• Category Trends:
  • C1 share declines significantly (-2.5 percentage points/decade, $P=0.001$).
  • C3 and C4 shares increase significantly (+1.0 to +1.2 percentage points/decade, $P < 0.05$).
  • C5 share shows a non-significant decline.

D. Basin Decomposition

• North Atlantic: Shows the strongest trend ($\Delta R = +0.126$), driven by both an increase in weak storms and a much faster increase in strong storms.
• Southern Indian: Shows a "pure intensification" pattern where weak storms are flat, but strong storms rise.
• Western Pacific: The largest basin by observation count shows essentially no trend.
• Eastern/South Pacific: All categories decline, but weak storms decline faster, artificially raising the ratio.

5. Significance and Conclusion

The paper concludes that the narrative regarding tropical cyclone intensification has evolved:

• 1979–2017: The observed intensification was largely a statistical artifact of fewer weak storms (C1) being recorded relative to the total.
• 1979–2023: The signal is now genuine. The data supports a physical shift where the frequency of major cyclones (C3 and C4) is increasing, even as the frequency of weak cyclones (C1) decreases.

This distinction is crucial for climate science. It suggests that while detection issues or environmental factors may still suppress the count of marginal cyclones, the thermodynamic environment is now robustly supporting the formation and maintenance of high-intensity tropical cyclones. The study validates the Kossin et al. hypothesis but refines the mechanism, confirming that the intensification is no longer just a "denominator effect" but a "numerator effect" driven by more frequent major storms.