Weather Jiu-Jitsu Shifts Sandy 322 km, But Only in a Model

Estimated reading time: 10 minutes

Key Takeaways

• 322 km, in a model only: In a simulation, a nudge applied seven days before landfall moved Hurricane Sandy's track by about 322 km (roughly 200 miles). It was not a field test (PLOS Water essay, June 24, 2026).
• Small inputs, idealized models: The paper reports control inputs below 2% of total system energy in idealized models. The Sandy, Texas and California runs are separate simulations.
• WMO pushes back: The World Meteorological Organization says there is no generally accepted evidence that tropical cyclones can be modified (WMO weather-modification statement).
• Two other simulated results: The 2021 Texas freeze run cut the minimum-temperature anomaly by about 10°C. The 2022 atmospheric-river run cut integrated vapor transport by about 5%.
• Not tested in real conditions: The Debrief reported on June 25, 2026 that the work had not been tested in the real world. PLOS published a correction to nine references on August 20, 2026.

The PLOS Water paper behind weather jiu-jitsu reports that its simulated control inputs stayed below 2% of total system energy in idealized models. Yet the World Meteorological Organization (WMO) says there is no generally accepted evidence that tropical cyclones can be modified. Both statements can be true at once, and the gap between them is the story.

In one simulation, the authors shifted Hurricane Sandy's track by about 322 km. They applied a nudge seven days before landfall. It happened only inside a computer model of the 2012 storm.

Why should founders and operators care? The proposal hints at a possible long-term climate-risk technology opportunity. But its present status means investment, insurance, liability and regulatory decisions should not treat simulated track changes as an operating capability. This post separates what was shown from what was not.

How the Aurora AI Weather Model Was Used to Find Where to Push

The paper is a perspective essay in PLOS Water, published on June 24, 2026. Its authors are Qin Huang, Moyan Liu and Upmanu Lall. Gizmodo's article by Ellyn Lapointe, published the same day, describes Huang as a PhD student at Arizona State University.

The core idea is simple. The jet stream, the fast river of air that steers weather, is unstable in predictable ways. Huang told Gizmodo the team wanted to ask whether that instability could be used: "a small nudge at the right time and place, amplified by the atmosphere's own dynamics."

What the model did

The team used two main tools, according to the PLOS Water essay:

• Finite-Time Lyapunov Exponent (FTLE) diagnostics. This is a math tool that shows where air flow pulls apart fastest, so a small change there can grow.
• Aurora. This is a deep-learning Earth-system model. Gizmodo calls it a large-scale AI model built for high-resolution weather prediction. It is the aurora ai weather model at the center of the experiments.

The authors applied a one-time perturbation at two or three locations and atmospheric levels. A perturbation is a small push added to the weather system.

How big the nudge was

The simulated input matched a temperature anomaly of about 6.7–10.4 K and a specific-humidity anomaly of about 2.3–3.4 g/kg. A kelvin (K) is the same size as a degree Celsius when you measure a change. Specific humidity is the amount of water vapor per kilogram of air.

The paper separately reports control inputs below 2% of total system energy in idealized models. Our research does not explain how that figure relates to the anomalies above, so treat them as different measures.

All of this is author-reported simulation. It is not a field demonstration.

Gizmodo says the delivery idea is cloud seeding. Traditional seeding adds particles to the lower atmosphere to increase local rain or snow. Here, the goal would be a small disturbance days before an extreme event. "We are not trying to change what happens at the seeding site," Huang said. "We are trying to trigger a cascade that reshapes a weather system hundreds or thousands of miles away."

Source: PLOS Water essay, June 24, 2026. Quotes and the cloud-seeding description come from Gizmodo's June 24, 2026 article by Ellyn Lapointe.

The Hurricane Sandy Track Shift: 322 km, With a Catch

Hurricane Sandy is a useful test case because the damage was so large. Gizmodo reports that about 60 million people across 24 states felt its effects in October 2012. It also reports that New York and New Jersey saw extreme storm surge and more than $62 billion in damages.

In the simulation, perturbations applied seven days before landfall shifted the storm's track by about 322 km. That is the hurricane Sandy track shift in the headline. Gizmodo says the modeling suggests the storm stayed primarily offshore.

The paper does not show that a real hurricane could be moved that far. The result is described as a model-based proof of concept.

The catch MIT's Kerry Emanuel flagged

Kerry Emanuel, a post-tenure professor of atmospheric science at MIT, told Gizmodo that the modified track still came dangerously close to Cape Cod, Nantucket and Martha's Vineyard. Moving a storm away from one place can put it near another.

Gizmodo's article also carries a pull quote on the same worry: "If you steer it away from New Jersey and it destroys Nantucket, you're going to be in for some lawsuits." The article does not name the speaker, so we do not attribute it.

Emanuel also described a trade-off between energy and predictability. "The less energy you have to put into the atmosphere, the longer you have to let the perturbation grow," he said. That means seeding earlier and getting a less predictable outcome. He added that where the best balance lies is something he is "not sure" about.

What the authors say

Huang said the goal was "to show that a small perturbation produces a measurable track shift." She added that a real system would need a better objective: continuously steering the storm toward the open ocean, with its path re-checked and corrected at each step.

Source: PLOS Water essay, June 24, 2026. Sandy impact figures, the 200-mile conversion and the Emanuel and Huang quotes come from Gizmodo's June 24, 2026 article by Ellyn Lapointe.

Texas Freeze and Atmospheric River: Two More Simulated Results

The team ran two other case studies. Each was a separate simulation, and none is a measured weather outcome.

The 2021 Texas freeze

The paper describes a reduction of the minimum-temperature anomaly by about 10°C. An anomaly is how far the temperature sits from normal. Gizmodo put the result as raising minimum temperatures by about 18°F (10°C).

The 2022 California atmospheric river

An atmospheric river is a long, narrow band of moisture in the sky. This run reduced integrated vapor transport by about 5%. Integrated vapor transport is the total water vapor moving along in the air.

The paper ties the result to a one-time perturbation under favorable upstream kinematic conditions. In plain words, the air flow upstream was set up in a way that helped the nudge work. Gizmodo says the nudge also shifted the river's trajectory and tied the event to flooding in California in 2022.

Why the three results should stay separate

• Each event had its own setup and its own nudge.
• A 5% change in vapor transport is not a 5% change in flood damage.
• A model temperature shift does not mean a real freeze could be softened.

The authors present all three as a first proof of concept, not an optimized modification experiment.

Source: PLOS Water essay, June 24, 2026. The Fahrenheit conversion and the California flooding context come from Gizmodo's June 24, 2026 article.

Cloud Seeding Hurricane Control Meets the WMO's Skeptical Line

Cloud seeding is the delivery idea Gizmodo links to weather jiu-jitsu. The WMO's statement is the strongest independent expert-status reference in our research. It was first published on June 14, 2025, and the page is marked as last updated July 2, 2026. It predates the PLOS paper, so it should not be read as a reply to it.

The WMO statement makes these points:

• Weather modification is generally local to regional.
• Its effectiveness depends on cloud and atmospheric conditions.
• There is no generally accepted evidence that tropical cyclones can be modified.
• There are no demonstrated cloud-seeding methods that modify floods or other severe weather.
• Large-scale claims about altering wind patterns to bring water to a region, or eliminating severe weather, lack a sound scientific basis.

Our research summary also says the WMO supports further scientifically designed research. The statement separates local-to-regional cloud seeding from climate intervention.

Where the paper and the WMO sit

The paper is a simulation essay. The WMO is describing what has been demonstrated in practice. These are different kinds of evidence, and the paper does not claim a field demonstration.

Emanuel adds a practical doubt. He told Gizmodo he is skeptical that cloud seeding is the best way to create these perturbations, because it only works under certain meteorological conditions.

Is there an expert consensus?

No. The Debrief reported on June 25, 2026 that the work had not been tested in real-world conditions. Beyond the WMO position and Emanuel's comments in Gizmodo, our research found no additional independent technical commentary. A consensus supporting operational hurricane steering cannot be established from the evidence available.

Source: WMO weather-modification statement; The Debrief report, June 25, 2026. Emanuel's comments come from Gizmodo's June 24, 2026 article.

A Three-Stage Research Plan, Not a Launch Date

The Debrief reports that the authors propose a three-stage path:

1. More simulations and observation design. This means further deep-learning perturbation experiments and designing the observation systems needed to track results.
2. Regional pilot studies and governance work. This stage involves the WMO and UNEP, the United Nations Environment Programme.
3. Limited operational pilots under formal international oversight.

The Debrief presents this as a research agenda. It is not an announced deployment schedule, and no dates are attached.

What would have to improve first

The research team says real-world use would need advances in monitoring, attribution, delivery methods, evaluation and international governance. Attribution means proving that a change in the weather was caused by the nudge and not by chance.

Huang told Gizmodo the longer-term vision includes field experiments in safe settings. One example is trying to steer a storm entirely over the open ocean, far from any populated coast. She called that the bridge between simulation and any real operational use.

The legal reference

The paper cites the United Nations Convention on the Prohibition of Military or Any Other Hostile Use of Environmental Modification Techniques as a governance reference. Gizmodo says the authors also acknowledge technical, legal, social and environmental challenges that future studies must address.

Source: The Debrief report, June 25, 2026; PLOS Water essay, June 24, 2026. The field-experiment quote comes from Gizmodo's June 24, 2026 article.

No Funding, a Conference Listing and a Nine-Reference Correction

Some paper-trail details help when judging how much weight to give the work.

• Funding and conflicts. The paper reports no specific funding and no competing interests. That is an author disclosure. It tells you no company is behind the work, but it also says nothing about the quality of the results.
• A conference listing. The timeline also records a European Geosciences Union (EGU) General Assembly listing from May 2026. It presented weather jiu-jitsu as an atmospheric-nudging proposal and reported perturbation experiments across tropical cyclones, atmospheric rivers and freezes. See the EGU General Assembly listing. A conference abstract is not the same as a reviewed field result.
• A correction. On August 20, 2026, PLOS published a correction addressing errors in nine references. Our research found no reported change to the three headline simulation results.

A reference correction is routine housekeeping, but it is worth knowing about if you cite the paper in a pitch deck or a risk memo. Check the corrected version first.

Source: PLOS Water essay, June 24, 2026; PLOS Water correction, August 20, 2026; EGU General Assembly listing.

What This Means for Startups and Small Businesses

This is our analysis, based on the evidence above. Weather jiu-jitsu is a research proposal. Here is how to treat it in practice.

• Do not price it in. A simulated 322 km track shift is not an operating capability. Leave it out of financial models, insurance assumptions and customer promises.
• Watch for stage markers. The meaningful signals are the ones in the authors' plan: more experiments, regional pilots, WMO and UNEP governance work, and open-ocean field tests. None of these are scheduled.
• Plan for liability early. Emanuel's point about people "adversely affected" and the lawsuit warning in Gizmodo's article show the issue. If a storm is steered away from one place, someone else may bear the risk. Insurers and legal teams will ask who is responsible.
• Expect regulation to lead. The paper cites an international treaty on environmental modification, and the WMO says large-scale claims lack a sound scientific basis. Any product built near this field would face close scrutiny.
• Value the skills, not the storm control. Simulation, AI forecasting and risk modeling are useful now. A climate-risk team can use them for scenario planning without claiming to change the weather.
• Check the status of every claim. Ask whether a number comes from a model, a field test or a plan. The same 322 km figure means very different things in each case.

For most small businesses, the practical move is boring and sound. Prepare for extreme weather with backups, remote work options and clear contracts. Do not wait for anyone to steer storms away.

Frequently Asked Questions