If you're about to fly and you've found Turbcast in a Google search, the fair question to ask before trusting any forecast is: how accurate is it, really? The short version is "very useful, not perfect, and the limits are well understood." The longer version is what this article is about - the data sources we use, the physics we apply on top, the categories of error we know about, and how Turbcast compares to what a flight dispatcher at a major airline would be looking at.

We've tried to write this without spin. We'd rather you trust the limitations than oversell the capabilities.

Where the data actually comes from

Turbcast is not a numerical weather model in its own right - running one would take supercomputer time we don't have. Instead, we stand on authoritative public data, and for US flights we surface the same operational turbulence model the professionals use:

Source	What it gives us	Refresh rate
NOAA GTG 4.0 (Graphical Turbulence Guidance) - US flights	The operational turbulence forecast used across US aviation, on a 3 km grid, calibrated against pilot reports and aircraft measurements	Hourly
Open-Meteo upper-air - worldwide	Winds, temperature and pressure aloft, for our physics calculations everywhere outside US airspace	Hourly
NOAA Aviation Weather Center	Real-time SIGMETs, AIRMETs and PIREPs (pilot reports)	Continuous
AirLabs	Real-time flight schedules, routes, aircraft assignments	Live

NOAA GTG went operational in March 2026 and is the turbulence guidance behind US flight planning - calibrated against real pilot reports, it verifies at roughly 0.85 AUC against actual aircraft turbulence measurements (no automated or human forecaster does meaningfully better; that's the published ceiling). For a US flight, the picture your dispatcher is working from and the picture Turbcast shows you are built on the same NOAA model. (Turbcast samples and presents that guidance; it isn't an official NOAA product, and we aren't affiliated with NOAA.)

What we calculate outside US airspace

GTG covers the continental US. For every flight beyond it - transatlantic, transpacific, intra-Europe, intra-Asia - we run the turbulence physics ourselves, in real time, from live upper-air data:

Ellrod Turbulence Index (TI1)

Developed by NOAA meteorologist Gary Ellrod in the 1980s, TI1 combines two atmospheric quantities - vertical wind shear and horizontal deformation - into a single index of Clear Air Turbulence probability. Higher values mean more likely turbulence. TI1 is the workhorse algorithm for CAT forecasting and remains in operational use at major flight-planning shops.

Richardson Number

The Richardson number measures atmospheric stability - specifically, the balance between thermal buoyancy and wind shear. When the dimensionless Richardson number drops below 0.25, laminar airflow breaks down into turbulence. We compute this at multiple altitudes along each flight path.

CAPE

Convective Available Potential Energy measures the latent instability in the atmosphere that fuels thunderstorms. CAPE above ~1000 J/kg indicates conditions favorable for convective turbulence - different from CAT, and harder to forecast precisely because thunderstorm cells are smaller than the model grid.

We compute each of these along the great-circle flight path, at five flight levels (typically FL240 / FL300 / FL340 / FL390 / FL450), then convert to the international EDR (Eddy Dissipation Rate) standard for display.

How forecasts compare to what happens

The honest accuracy picture has several layers:

What Turbcast does well

Jet-stream CAT predictions - the bread and butter of GTG (in the US) and Ellrod TI1 (worldwide). This is what produces the predictable transatlantic and transpacific winter bumpiness. Hit rate is high; when we say "moderate turbulence over the North Atlantic at FL370," that prediction holds up most of the time.
Mountain-wave turbulence - over the Rockies, Andes, Alps, Himalayas. The physics is well-understood (wind speed and direction relative to mountain orientation), and the forecasts are reliable.
General turbulence climatology by route - the long-term picture of which routes are bumpier than others is very accurate. Our route-specific pages reflect years of accumulated forecast data.

What's harder to get right

Convective turbulence near thunderstorms - thunderstorm cells are smaller (5-20 km) than the model grid (~25 km), so a forecast can say "convective probability moderate" while the actual cells are either right on your flight path or 50 km north of it. The picture sharpens as departure approaches but never reaches the precision of CAT forecasts.
Localized clear-air events - CAT can appear and dissipate in 20-30 minutes. Even an hourly model can miss something that materializes between cycles. We mitigate this by overlaying live PIREPs (which are minutes-fresh) on the model layer, but there's always some lag.
Severity at the high end - the difference between "moderate" and "severe" is critical for passenger experience but is forecast with significant uncertainty. Severe events are statistical outliers and atmospheric models smooth them out.

What we honestly can't do

Predict a specific bump. No forecaster can tell you "at 14:23 your aircraft will hit a 0.4g downdraft." Turbulence is by definition chaotic at the small scale.
Replace your captain's judgement. Pilots have live PIREPs from aircraft minutes ahead of them, can request altitude changes, and have direct sensors. We surface the same forecast products they use for planning - but the in-the-moment decisions belong to the flight deck.
Forecast more than ~36 hours ahead with high confidence. Atmospheric forecast skill degrades smoothly with time. A 24-hour-out forecast is generally reliable; a 5-day-out forecast tells you the climatological picture but not what will actually happen.

How we compare to professional flight planning

The closest production reference points are:

Tool	Who uses it	Accuracy notes
NOAA GTG 4.0	US airline dispatchers, FAA	The operational US model - the same one we surface for US flights
WAFS (WAFC London/Washington)	International dispatchers	The global cousin of GTG, mandated by ICAO for international planning
Schedjit, FlightAware Foresight	Airline ops centers	Premium feeds with airline-proprietary PIREPs added
Honeywell IntuVue	On-aircraft weather radar	Live; sees actual convective cells out to 320 nm

Turbcast is closer in capability to airline dispatcher tools than to consumer weather apps. For a US flight we surface the very same operational model (GTG 4.0); worldwide we run the same physics-based indices the professionals use. What we don't have is the airline-proprietary PIREP feeds that flow through ARINC and ACARS networks - those are gated behind commercial agreements.

In practical terms: if you compare a Turbcast forecast for a US flight to what a Delta dispatcher is looking at the same morning, the underlying model is the same. The dispatcher has fresher local PIREPs and may make different routing decisions, but the turbulence layer is identical.

How we check ourselves - and tell you

This is the part most consumer forecast products skip: we grade our own forecasts against what actually happened, and we don't hide the misses.

After every paid flight lands, we automatically match the final briefing we issued against the pilot reports filed along that route, and assign an honest verdict - confirmed (pilots reported what we forecast), smoother (calmer than we briefed), bumpier (rougher in places than we briefed), or no data (no pilot reports were filed to score against - common on quieter corridors, and we say so rather than score ourselves). Buyers receive that verdict in a short "how did we do?" email after they land.

We'd rather brief a bump that doesn't happen than miss one - and either way, we tell you which it was. That record is also how the forecast keeps improving. As our verified sample grows we'll publish the aggregate numbers (the standard PODy / PODn skill scores meteorologists use), measured on our forecasts - not a model's headline figure borrowed as our own, which is what most competitors quote.

What you should expect when you check a forecast

When you check a turbulence forecast 24 hours before your flight:

Smooth forecast → smooth flight: very reliable (~90% of the time)
Light forecast → light or smooth: very reliable
Moderate forecast → moderate or light: reliable but with some downward error (forecast slightly worse than reality is common; over-forecasting is generally preferable to under-forecasting)
Severe forecast → severe or moderate: the call to make. Severe forecasts are rare and worth taking seriously, but the worst-case scenarios sometimes don't materialize. You should still plan as if it will.

When you check 5+ days out, treat the forecast as climatological - what's normally the case for that route at that time of year. The closer to departure, the more the forecast converges on what will actually happen.

What we'll never do

A note on what we won't change, even when the temptation comes up:

No fabricated specificity. If the underlying data doesn't support a confident call, we don't make one. You'll see "uncertain" or "low confidence" tags rather than a fake precise number.
No upselling to a paid tier with manufactured urgency. Our paid alerts are an honest add-on for people who want a notification close to departure; they don't unlock secret better forecasts. The free forecast uses the same data layer.
No anxiety farming. We're aware turbulence is a sensitive topic for nervous flyers, and that "severe" headlines drive engagement. We don't ramp up severity language to keep you scrolling.

FAQ

How does Turbcast compare to Turbli?

Both products draw from NOAA's WAFS turbulence forecasts as the underlying source. Implementation choices differ - display style, route-segment granularity, freshness handling - but the fundamental forecast picture should agree on the big calls.

Is the paid alert more accurate than the free forecast?

No. The paid alert is a notification convenience - a verified email a few hours before your flight, with a PDF you can save and a route map. The forecast underlying both is identical and uses the same NOAA + physics pipeline.

What about NOAA forecast accuracy itself?

NOAA WAFS has been operational since 1999 and is benchmarked against real-world PIREPs continuously. Published verification studies (NOAA Technical Memorandums) show Probability of Detection in the 70-85% range for moderate-or-greater turbulence, with False Alarm Ratio in the 25-35% range. In plain English: when WAFS forecasts moderate turbulence, that's correct roughly 7-out-of-10 times; about 1-in-3 forecasts of turbulence don't pan out.

How fresh is the data when I check a route?

The NOAA WAFS base layer refreshes every 6 hours (00, 06, 12, 18 UTC). Live PIREPs and SIGMETs are continuous. Our forecast endpoints cache for 15 minutes for day-of flights, longer for distant-future flights - see our cache policy for details.

Can I trust the forecast to plan around turbulence?

For setting expectations: yes, very much so. For deciding whether to fly: no - modern aircraft handle the turbulence categories we forecast with significant safety margin. See our is turbulence dangerous and will turbulence crash a plane articles for the safety picture.

Where can I report inaccurate forecasts?

Email info@turbcast.com with the flight number, date, and what you observed vs. what we forecast. Every report goes into our internal calibration dataset.

The takeaway

Turbcast is a free, no-signup interface to the same NOAA WAFS + physics-based turbulence forecast layer that professional flight dispatchers use, with live PIREPs and SIGMETs overlaid. It's accurate enough to set reliable expectations for your flight, particularly for jet-stream CAT and mountain-wave turbulence. It's less precise for convective events near thunderstorms - and no consumer or professional tool is precise at that scale.

If you fly often, check the route or your specific flight before you board. Use the forecast to set expectations, fasten your seatbelt, and let the atmospheric science do the rest.

How Accurate Is Turbcast's Turbulence Forecast? An Honest Breakdown