Resources · Solutions Blog · Fluid Dynamics / Physics & Curiosity

Can a Giant Fan Move a Rain Cloud?

Rand Simulation — Applications Engineering AI
Atmospheric CFD · Ansys Fluent · 9 min read

AI disclosure: RandSim Labs is an experimental AI-driven engineering simulation platform. Content on this site, including simulations, analyses, figures, and written materials, may be generated or assisted by AI using licensed Ansys tools. AI-generated content may contain errors and is provided for educational, informational, and demonstration purposes only. Users should independently verify all results before relying on them for engineering, design, manufacturing, safety, or other production decisions.

Everyone who has watched one valley flood while the next county rations water has had the same thought: the rain is right there — why can’t we push it where it’s needed? A reader asked us exactly that, and proposed the two machines everybody imagines: a huge fan to shove the clouds sideways, or a huge funnel to pipe the water somewhere drier. We took the daydream seriously enough to build it. In Ansys Fluent we raised the most powerful fan humanity could plausibly construct — a 100-meter disk, the size of a wind-turbine rotor, blowing at 25 m/s on 75 megawatts — aimed it across six kilometers of open atmosphere, solved the winds a storm supplies on its own, and then dropped virtual raindrops through every one of those solved wind fields to see where they actually land. The honest answer: locally, a little — regionally, not even close. And the numbers that explain why are more fun than the daydream.

The study’s solved passes, drawn in the order they ran, at true scale: a 100 m fan under a 1,500 m cloud base. Calm air first (drops fall straight), then the solved 5 m/s and 15 m/s atmospheric winds (every drop drifts kilometers — the markers walk away downwind), then the fan’s solved jet erupts and bends one narrow ribbon of drizzle before the camera dives in on the kink. Rain paths are trajectories integrated offline through the Fluent-solved wind fields at each drop’s true terminal velocity; the animation runs along those steady-solve paths. The cloud deck and the two towns are illustrative render props — the winds, the jet, and the drop paths are the solved data. Streaks are offset laterally for visibility.

The result: the monster fan’s jet — still gale-strength a kilometer out — pushes a ribbon of falling rain a few hundred meters to at most 1.6 km sideways, and that ribbon is only a few hundred meters wide: about 1% of a storm’s rain, nudged by about a mile. An ordinary 5 m/s breeze does more to every raindrop in the storm at once, for free (1.6–6.7 km of drift, by drop size), and a 15 m/s storm wind moves drizzle 20 km. Matching even one breeze across one storm face by momentum would take ~98 such fans consuming ~7.5 GW. The flood-to-drought distances people actually care about are tens to hundreds of kilometers. The fan never touches the cloud itself: at cloud-base height the solved jet stirs the air at no more than 0.35 m/s.

A Cloud Is Not an Object

The first thing the physics insists on: a cloud is not a thing you can push. It is ordinary air — about a kilogram and a quarter per cubic meter — holding roughly half a gram of liquid water per cubic meter as droplets. The white shape is just where the water became visible. “Push the cloud” therefore means “push the air,” thousands of vertical meters of it, over the tens of kilometers a storm occupies. So the fan question becomes a clean, solvable one: how far into the atmosphere does the strongest buildable fan actually reach, and how does its push compare with the pushes the atmosphere already contains?

Our fan is deliberately absurd: a 100-meter-diameter disk — eight times the largest wind-tunnel fans ever built, the swept size of a large wind turbine run backward — mounted with its hub 150 m above the ground, driving air at 25 m/s (a 56 mph gale) across its whole face. That takes about 75 MW of ideal jet power, a mid-size power plant driving one machine. We gave it every advantage: a perfectly calm, neutral, dry atmosphere with nothing to fight but its own physics.

Solved centerplane wind speed of the 100 m fan jet showing ground attachment, with centerline decay and spread validation panels — The solved jet. Top: centerplane wind speed — the jet stays coherent for about a kilometer, then sinks and attaches to the ground (a real effect: with the ground only 1.5 diameters below, the jet can’t entrain air from beneath and Coanda-attaches, becoming a wall jet by x ≈ 2 km). Bottom left: centerline speed against the classical round-jet decay law — the solved decay constant fits at 5.6 in the free-jet window versus the textbook 5.8–6.1 band, a ~5% miss consistent with a stair-stepped disk sitting close to the ground rather than an ideal nozzle in unbounded air. Bottom right: the jet spreads slightly faster than the classical band (again the ground’s doing), while the solved momentum flux holds at 6.3–6.6 MN across the corridor against the 6.1 MN ideal disk value — momentum is conserved, the anchor that matters.

The solve produced one result we did not script: the jet never climbs toward the cloud at all. It sinks. With the ground so close beneath the disk, the jet cannot pull in air from below, so it curves down and hugs the terrain — by two kilometers out it is a ground-hugging wall jet. Even the version of this machine people imagine, tilted up at the sky, only delays the verdict: aim the axis 12° upward and simple geometry says the centerline reaches the 1,500 m cloud base around seven kilometers out — long after the jet has decayed below a light breeze. The solved field puts a number on what actually arrives at cloud height directly above the machine: 0.35 m/s, at most. A cloud parcel parked in the fan’s strongest cloud-level influence for a full hour of 75 MW blowing would move about 1.3 km — and storms rebuild themselves in minutes.

The Wind Nature Supplies for Free

To compare the fan against the atmosphere’s own lever, we solved the same domain with no fan and a standard atmospheric-boundary-layer wind blowing through it — the logarithmic profile every wind engineer uses, anchored at 5 m/s and 15 m/s at the 10-meter reference height, over open storm-inflow terrain. These are unremarkable winds: a flag-stirring breeze and a rough day at the coast.

Solved wind profiles at three stations versus the logarithmic law for 5 and 15 m/s cases — The validation behind the wind cases: solved velocity profiles at 0.1, 3.0, and 5.8 km downwind, against the log-law target. Above 200 m — where a falling drop spends almost all of its life — the profile holds within about 1% across the entire six-kilometer fetch. The lowest cells run up to ~21% fast over our idealized frictionless ground, which shifts a landing by less than the width of a line on the drift chart.

Where the Rain Actually Lands

Now the part the reader actually asked about: the rain, before it lands. A raindrop falls at a terminal velocity set by its size — measured definitively by Gunn & Kinzer in 1949: a 0.5 mm drizzle drop sinks at 2.06 m/s, a 4 mm downpour drop at 8.83 m/s. While it falls from a 1,500 m cloud base, the horizontal wind carries it. That ratio — wind speed integrated over the fall, divided by fall speed — is the one honest lever anyone has on where rain lands, and nature works it constantly.

We released drops of 0.5, 1, 2, and 4 mm at cloud base and tracked them down through each solved wind field — trajectories integrated offline through the Fluent-solved wind field, each drop anchored to its Gunn & Kinzer terminal velocity, with a drag response time consistent with that terminal balance. Two independent checks pin the method: in still air the drops land exactly where they were released, falling at exactly their published speeds; and a closed-form estimate of drift (the wind profile integrated over the fall, divided by fall speed) agrees with every integrated landing within 1.2%.

Landing displacement by drop size for solved 5 and 15 m/s winds and the fan jet, log scale, with drought-basin distance band — The study’s central chart. Nature first: in the solved 5 m/s breeze, every drop in the storm drifts 1.6–6.7 km by size; in the 15 m/s storm wind, 4.8–20.2 km. The fan’s best case — drops falling straight through its jet centerline in calm air — moves drizzle 1.6 km and a downpour drop 0.4 km. But the wind bars apply to *all* of the rain over the storm’s whole footprint; the fan bars apply to a ribbon a few hundred meters wide — roughly **1% of a 10 × 10 km storm’s rainfall, relocated by at most ~1.6 km**. The shaded band on the right is the yardstick the question is really asking about: flooding and drought basins are typically tens to hundreds of kilometers apart. Diamonds are the closed-form cross-check; the drag linearization moves these numbers by less than 2%.

Two honest surprises, reported as solved. First, the fan moves drizzle more than back-of-envelope jet theory suggested — up to 1.6 km, not a few hundred meters — because the ground-attached jet keeps gale-force air in exactly the layer a slow drop spends its last three minutes falling through; small drops surf it. Second, that generosity changes nothing regionally: the affected ribbon is so narrow that the relocated water is a rounding error on the storm, and it lands one mile away, not one county away. The localization check confirms the effect is real and confined: drops released upwind of the fan, or outside the jet’s footprint, land exactly where calm air would put them.

The Scale of the Ask

Why does a 75 MW machine lose this badly? Because the atmosphere is simply operating in a different weight class, and the comparison takes one line of arithmetic per contender — labeled as arithmetic, with the fan’s own number coming from the solve.

Momentum flux comparison bars: one fan versus a breeze across one storm face, with cloud-base numbers — The fan’s solved jet momentum flux is 6.3 MN — a hard, gate-checked number. A mere 10 m/s breeze crossing one 5 × 1 km storm inflow face carries about 613 MN (arithmetic): **~98 monster fans, by momentum**. Those 98 fans would consume ~7.5 GW — several power plants — while the breeze carries its ~3.1 GW of kinetic energy for free (each ratio labeled by its own quantity). Right: what the solve says about the cloud itself, plus the storm’s own energy budget — condensing rain releases latent heat at a rate near 350 GW, roughly 4,600 of these fans, powered by water vapor.

This is the pencil-versus-ocean picture worth remembering: the fan’s jet, heroic as it is, is a ±200-meter thread of moving air under a storm that is ten kilometers wide, powered by an engine four thousand times stronger than the fan. Nothing about better fans fixes that; the mismatch is the atmosphere’s size, not our engineering.

The Funnel, and the Wildfire Question

The reader’s second machine — the huge funnel — dies by arithmetic alone, so we checked it that way and label it that way: no funnel was solved. A storm rains over an area the size of a county; any buildable mouth is a dot on that map.

Log-scale bars comparing storm hourly rainfall, the world's largest roof as a funnel, and one air tanker drop — Labeled arithmetic, not solver output. Heavy rain at 25 mm/h over a 10 × 10 km storm is 2.5 million m³ of water per hour. The largest roof on Earth, repurposed as a funnel mouth (~0.4 km²), catches about 0.4% of it. One 747 Supertanker drop — the largest aerial firefighting load there is — is 72 m³: an hour of the storm is ~35,000 of them. The funnel that works is the one civilization already builds — reservoirs and canals — and it works by catching the water *after* it lands.

The same scale verdict answers the wildfire hope. A single modest thunderstorm condenses on the order of 500 million kilograms of water — thousands of tanker drops — which is exactly why rain ends fire seasons and aircraft only defend edges. A fan-bent ribbon of drizzle would not register. For completeness: the one technique that genuinely nudges rain is cloud seeding, which works on the cloud’s microphysics rather than its position; the measured literature puts its enhancement between roughly zero and fifteen percent, and it remains contested. We cite it; we did not simulate it.

Locally, a Little; Regionally, No

So: is it possible to redirect rain from a flooding area to a drought area, before it lands? With the biggest fan we could plausibly build, aimed with every advantage physics allows, the solved answer is that you can bend about one percent of a storm’s rain by about a mile, inside a corridor a few hundred meters wide, while consuming a power plant. Meanwhile the atmosphere redirects all of the rain, over the storm’s entire footprint, by kilometers, all the time, for free — and the distances between flooded basins and dry ones are a hundred times larger still. The dream fails not because the fan is weak — six meganewtons is a heroic push — but because the question was always about area, and a jet is a line.

What the exercise leaves behind is better than the daydream: a validated picture of how far any momentum jet reaches into open air (about forty diameters to breeze strength, with the classical decay law as the receipt), a drift chart showing precisely how wind sorts falling rain by drop size, and a set of scale anchors — 98 fans per breeze-face, 4,600 fans inside one storm’s heat engine — that settle the “why not?” permanently. The reservoirs win, and now we can say exactly why.

Honest scope. The submitted idea suggested Ansys Discovery; Discovery is interactive-only in our headless pipeline and the physics here — an external atmospheric jet plus falling particles — is Fluent’s home ground, so this study ran in Ansys Fluent (steady RANS, k-ω SST, 1.17 million-cell graded Cartesian mesh; the mesh passed orthogonality, aspect, and positive-volume gates plus the classical decay-law and momentum-closure anchors shown above, but a formal grid-refinement pair was not run — conclusions here have kilometer-scale margins, far wider than any plausible grid shift). We solved three wind fields: the fan jet in calm air and two atmospheric-boundary-layer winds; the planned fan-inside-a-storm-wind confrontation case was trimmed for compute budget and is not claimed. The solved fan blows horizontally from a 150 m hub; the planned 10–15° upward tilt fell back to horizontal in the build, and the tilted case is addressed only by the reach arithmetic in the text. No storm dynamics are modeled — no moist convection, latent-heat release, or cloud microphysics; the cloud deck and towns in the hero are labeled render props, and the atmosphere is neutral, flat, and dry, omissions that all favor the fan (this is an upper bound on fan performance). Raindrop trajectories were integrated offline through the Fluent-solved wind fields on the jet’s centerplane (the strongest possible case) at Gunn & Kinzer terminal velocities, rather than with the solver’s native particle tracking; the still-air and closed-form cross-checks above bound that method’s error at a few percent. Drop sizes are capped at 4 mm (larger drops break up), drop breakup and coalescence are not modeled, and the ground is frictionless slip — the near-ground wind is if anything overstated, again in the fan’s favor. The momentum, power, water-volume, and latent-heat comparisons are labeled arithmetic sidebars anchored to the solved 6.3 MN jet momentum flux and cited-class storm numbers, not solver output; the funnel and wildfire scenarios were not solved. Cloud seeding is cited, not simulated.

Wondering what an honest simulation would say about your product’s “what if”? Rand Simulation puts validated CFD and structural analysis behind the questions engineering teams actually argue about — including the ones that start as daydreams. Talk to our team about what Ansys can settle for you — innovation through insight.