How It Works

A plain-English look at the science behind your turbulence forecast

It starts with really good weather data

Everything we do is built on top of NOAA's High-Resolution Rapid Refresh (HRRR) model. It's the best real-time weather model available for the US, and it's the same data that professional meteorologists and aviation forecasters rely on every day.

Here's what makes it so useful for turbulence:

• 3 km resolution: That's detailed enough to pick up on small-scale atmospheric waves and ripples that bigger, coarser models just can't see.
• Native vertical levels: We run our calculations on the model's original terrain-following layers (up to 50 of them) instead of smoothing things out onto standard pressure levels. This preserves a lot of detail, especially near mountains.
• Updated every hour: New forecasts come in hourly and extend up to 18 hours out. So as your flight gets closer, the data keeps getting fresher and more accurate.

We don't just check one thing

Turbulence is tricky. It can come from jet streams, mountain waves, wind shear, temperature gradients, and a bunch of other atmospheric processes. If you only look at one of those, you're going to miss things or get a lot of false alarms.

So instead of relying on a single formula, we run 13 independent turbulence calculations on the weather data. They fall into two groups:

Clear-Air Turbulence (CAT): 11 diagnostics

These are focused on turbulence that happens in clear skies, usually at cruising altitude near the jet stream. Each one looks at the atmosphere from a slightly different angle:

• Ellrod3: Looks at how quickly air is spreading apart or converging. This one has historically been the single best predictor of clear-air turbulence.
• Fronto: Picks up turbulence along upper-level weather fronts and jet stream boundaries.
• NGM1: Combines wind speed with wind deformation to find shear zones near the jet stream.
• DEFSQ: Focuses specifically on extreme wind deformation, which tends to show up during the worst clear-air events.
• VORTSQ: Measures how strongly air is spinning relative to how stable the atmosphere is. Lots of spin plus low stability means trouble.
• IAWIND: Catches wind accelerations that don't follow the expected large-scale flow. These imbalances near the jet can kick off turbulence.
• EDRLL: Estimates turbulence intensity directly from the atmospheric energy spectrum. It's more theoretical but adds a useful independent signal.
• RITW: Finds layers where the wind changes direction sharply as you go up. That directional shear is a strong turbulence trigger.
• CTSQ: Looks at sharp temperature contrasts in unstable air. Think of it as "how jagged is the temperature field?"
• WSQRI: Tracks vertical air motion (updrafts and downdrafts) in dynamically unstable layers.
• SIGW: Measures how erratically air is moving up and down. More variance means bumpier air.

Mountain Wave Turbulence (MWT): 2 diagnostics

When wind blows over mountains, it can create waves in the atmosphere that propagate up to cruising altitude. These two diagnostics are specifically designed to catch that:

• MWT2: Combines a terrain-driven forcing factor (how much the mountains are "pushing" the air) with temperature variability at altitude.
• MWT12: Similar idea, but pairs the terrain forcing with horizontal temperature gradients instead.

Both of these use a terrain-aware component that factors in mountain height, slope steepness, and low-level wind speed. They work especially well over the Rockies, Sierra Nevada, and Appalachians.

Converting everything to a common unit

Here's a problem: each of those 13 calculations produces its own number in its own units. A raw "Ellrod3 value" doesn't mean anything to a passenger, and you can't directly compare it to a "VORTSQ value."

So we convert all of them to Eddy Dissipation Rate (EDR), which is the international standard for measuring turbulence. EDR describes how much turbulent energy is being dissipated in the atmosphere, and it's the same metric that aircraft sensors measure in real time.

The conversion uses a methodology developed by Sharman & Pearson (2017) for the FAA. Each diagnostic gets statistically mapped to EDR using a distribution that was calibrated against millions of real turbulence measurements from commercial aircraft. So when we say "this area has an EDR of 0.2," that's tied directly to what an actual airplane would feel flying through it.

We calibrated it ourselves for the HRRR

One important detail: the original research by Sharman & Pearson was done using coarser weather models (13 km and 28 km grids). The HRRR runs at 3 km, which means it resolves much sharper gradients in the atmosphere.

You can't just take calibration numbers from a coarse model and apply them to a fine model. The numbers would be off. So we processed a multi-year archive of HRRR data to derive our own calibration parameters for each of the 13 diagnostics. This makes sure the EDR values stay physically accurate at this higher resolution.

Blending it all together

Once every diagnostic has been converted to EDR, we blend them in two steps:

1. Average within each group: The 11 clear-air diagnostics get averaged together, and the 2 mountain wave diagnostics get averaged together. This gives us a CAT score and an MWT score at every point on the map.
2. Take the higher one: At each point, we use whichever is larger. If clear-air turbulence is the bigger concern at that location, that's what shows up. If mountain waves are stronger, those take over.

We do it this way because clear-air and mountain wave turbulence are fundamentally different physical processes. Averaging them together would water down the signal. By taking the max, the dominant turbulence source always comes through clearly.

From weather grid to your flight score

Now we have a full 3D turbulence map of the atmosphere. The last step is mapping your specific flight through it.

We trace your route and sample the turbulence values at regular intervals along the way at your expected cruise altitude. Each sample point gets classified into a severity category:

• 90 to 100 (Smooth): EDR below 0.15. Basically nothing. Your coffee is safe.
• 75 to 89 (Light): EDR 0.15 to 0.22. Some light bumps here and there. The seatbelt sign might flick on for a bit.
• 50 to 74 (Moderate): EDR 0.22 to 0.34. You'll definitely feel this. Walking in the aisle gets tricky, and the crew might sit down.
• 15 to 49 (Severe): EDR 0.34 to 0.45. Significant bumps. Loose items can go flying. Keep that seatbelt on.
• 0 to 14 (Extreme): EDR above 0.45. This is rare and intense. Stay buckled in.

Each category has a weight (100 for smooth, 85 for light, 50 for moderate, 15 for severe, 0 for extreme), and your final score is the weighted average across your whole route. So a flight that's 80% smooth and 20% light turbulence would land around a 97. Pretty good.

What we can't predict

We want to be upfront about the limitations. Turbulence forecasting is really good these days, but the atmosphere is chaotic, and no model is perfect.

• US coverage only (for now): The high-resolution HRRR model covers the continental United States. For flights that go over water or outside the US, we fall back to lower-resolution global models.
• Pop-up thunderstorms: We capture large-scale instability well, but summer thunderstorms can pop up fast and create very localized turbulence that's hard for any model to nail down in advance.
• Altitude estimates: We estimate your cruise altitude based on route distance. If ATC sends you higher or lower than expected, the actual conditions could be a bit different from what we forecasted.

That said, by combining NOAA's best weather data with peer-reviewed science and our own calibration work, we think this is the most transparent and rigorous turbulence forecast available to everyday flyers. We're always working to make it better.

---