Post-Hoc Rationalization in Climate Science – Watts Up With That?

But only today did I become aware of his own post-mortem analysis. Mann’s reflection on the 2024 Atlantic hurricane season provides a fascinating case study in how climate forecasts go astray—and how their authors work to reframe misses as meaningful contributions. While the piece starts by heralding an “active, destructive, and deadly” hurricane season, its true focus is a far less successful story: the failure of Mann’s own predictions to match reality.

The Setup: Bold Forecasts, But a Lackluster Outcome

Mann opens with dramatic language about “unprecedented” hurricanes and their links to warming oceans. His piece asserts that human-caused warming is driving stronger storms, citing rapid intensification in specific cases like Hurricane Milton (which reached 185 mph winds) and even speculating about the need for a hypothetical “Category 6” classification.

Indeed Milton nearly breached the threshold of 192 mph sustained winds argued by one recent study to constitute a whole new “category six” caliber of hurricanes that has emerged in an era of unprecedented ocean warmth. A separate study estimated that the deadly flooding in the southeastern U.S. from hurricane Helene was increased by 50% by human-caused warming.

https://michaelmann.net/content/reflections-2024-atlantic-hurricane-season

However, these claims about stronger hurricanes don’t align with the latest findings from the IPCC. The 6th Assessment Report (AR6) explicitly states that evidence for long-term trends in hurricane frequency and intensity is inconclusive, with no detectable increase in global cyclone frequency and mixed regional trends. While short-term variability may yield extreme seasons, the broader picture contradicts Mann’s insistence that climate change is creating reliably stronger storms. His narrative relies on cherry-picking extreme cases rather than examining the full record.

This misalignment extends to his forecasts. Mann and his team predicted an “extremely active” hurricane season, with a most likely estimate of 33 named storms. Instead, the season produced just 18—or 19 if we generously include a borderline case. For perspective, this falls well below Mann’s projected range of 27–39 storms. Not exactly the “record-shattering” season anticipated.

The Excuses: Blame the Variables, Ignore the Model

Rather than owning the shortcomings of their forecasts, Mann and his colleagues turn to a familiar playbook: citing unpredictable variables to explain their miss. They blame the Madden-Julian Oscillation (MJO)—a well-known atmospheric cycle—for suppressing storm formation during the crucial July and August peak. Saharan dust outbreaks are also invoked as a contributing factor.

So there’s no real discrepancy when it comes to the latter half of the season. It was basically as active as predicted. The puzzle is why July and August were so quiet despite clearly favorable seasonal large-scale climate conditions. This is where one runs into complications with intraseasonal variability. Of particular relevance is the so-called Madden-Julian oscillation or simply “MJO” to its friends. The MJO is a roughly 40-50 day oscillation in the tropical atmospheric circulation which influences the location of convection, which shifts east and west over the course of a single 40-50 day cycle. When the center of convection coincides with the tropical Atlantic, conditions are more favorable for tropical cyclogenesis.

https://michaelmann.net/content/reflections-2024-atlantic-hurricane-season

While these factors undoubtedly play a role, they are hardly “unknowns.” Both phenomena have long been recognized as influential in tropical weather. Shouldn’t a sophisticated statistical model, touted as one of the “most accurate,” account for such routine variability? Mann’s reliance on these explanations reads less as scientific insight and more as retroactive justification.

Alternative Models: When One Forecast Fails, Pivot to Another

In a curious twist, Mann highlights an alternative model that predicted 19.9 storms—a remarkably close match to the observed total. Yet he admits that this model is generally less reliable than the one that failed spectacularly. Touting the success of a secondary, less accurate framework seems like an exercise in hedging bets rather than improving the science. If your “backup” prediction worked better, why not use it as the primary model going forward? This sleight-of-hand raises more questions than it answers.

There is one other noteworthy detail here. Our group makes an alternative forecast in which tropical sea surface temperature (SST) in the main development region (MDR) is replaced with what we call “relative SST”, defined as the difference between MDR SST and the average SST throughout the entire tropics, which some researchers have argued might be a better predictor of Atlantic hurricane activity. While our previous analyses have found that this alternative model yields less skillful predictions, it is notable that this year it yielded a much more accurate prediction of 19.9 +/- 4.5 total named storms that was remarkably close to the seasonal total.

Stronger Hurricanes? Not So Fast

A particularly bold claim in Mann’s reflection is that human-caused warming is directly responsible for the rapid intensification and increased destructiveness of storms. He points to Milton and Helene as evidence, even citing studies that argue for the necessity of a new “Category 6” hurricane classification.

However, this narrative overlooks the IPCC’s cautious stance on the issue. The IPCC AR6 finds only low confidence in detectable increases in hurricane intensity globally. While there is moderate confidence in increased peak wind speeds for a subset of tropical cyclones, these trends are not consistently observed across all ocean basins. Moreover, the number of global cyclones has either decreased or remained flat. Mann’s narrative about a world of ever-stronger storms is more rhetorical than rigorously scientific.

Here is the reality of Tropical Cyclone trends from Roger Pielke Jr, and Ryan Maue

“We have identified considerable interannual variability in the frequency of global hurricane landfalls; but within the resolution of the available data, our evidence does not support the presence of significant long-period global or individual basin linear trends for minor, major, or total hurricanes within the period(s) covered by the available quality data.”

https://rogerpielkejr.substack.com/p/just-the-facts-on-global-hurricanes

Broader Implications: Science or Spin?

Mann’s post inadvertently highlights a key issue in climate science: the gap between predictive confidence and observed outcomes. When forecasts fail, climate scientists often claim the climate system is no longer behaving predictably—implying their models are still right, just incapable of adapting to a “changing system.” This raises a critical question: If the system is behaving unpredictably, how can forecasts demand such unwavering trust, let alone justify sweeping climate policies?

Mann’s reflection ends with an ominous flourish: “There could well be far more unpleasant surprises in the greenhouse.” While such warnings are useful for grabbing headlines, they risk undermining public confidence when paired with unfulfilled predictions. If we’re not sure what to expect, how can anyone claim certainty about the impacts of climate change? This is less science and more speculative storytelling.

Final Thoughts: Missed Marks and Moving Goalposts

The 2024 Atlantic hurricane season wasn’t the catastrophe Mann’s models forecasted. While the storms that did form were destructive, the season as a whole fell far short of expectations. This raises a critical issue: When predictions fail, should we be refining models or doubling down on fear-driven narratives?

Mann’s reflections suggest the latter approach. His insistence that missed forecasts still “teach us important lessons” sidesteps the more obvious conclusion: overconfidence in flawed models does more harm than good. Rather than admitting these forecasts have limits, he suggests the rules of the climate system itself may be changing. It’s a convenient excuse but an unconvincing one.

For climate science to have any credibility, it needs to grapple honestly with its uncertainties—not reframe every miss as proof of an unknowable, ever-shifting reality. Until then, one can’t help but see this as another chapter in the growing credibility crisis of climate forecasting.