The Current Icing Product, the NCAR Algorithm Behind the Icing Forecast on Your EFB, and the Difference Between a Probability and a Guarantee at Altitude

The icing overlay on your EFB is generated by the Current Icing Product (CIP), a three-dimensional probabilistic model developed by the National Center for Atmospheric Research (NCAR) and updated hourly from real observations. Understanding what it actually outputs - and where it falls short - is one of the most practical pieces of meteorological literacy a pilot can have.

Why Did Icing Forecasts Need to Improve?

Before CIP, pilots relied on AIRMETs that might cover four states and 8,000 feet of altitude and tell you to expect “light to moderate icing in clouds.” Combined with voluntary pilot weather reports (PIREPs) that described conditions from minutes or hours earlier, the icing picture was broad, imprecise, and often outdated.

A series of accidents in the 1980s and 1990s forced the issue. The crash of American Eagle flight 4454 near Roselawn, Indiana in 1994 - a British Aerospace ATR 72 that encountered supercooled large droplets and lost control - was a watershed moment. That accident drove FAA funding for serious icing detection research, and the answer came out of Boulder, Colorado, where NCAR’s Aviation Applications Program had been doing atmospheric science since 1960.

CIP went operational in the early 2000s and fundamentally changed the precision of icing information available to pilots and dispatchers.

How Does the NCAR Algorithm Work?

CIP ingests data from multiple sources simultaneously:

• Radar data, showing where precipitation exists and something about vertical cloud structure
• Satellite imagery, including infrared channels that reveal cloud top temperatures
• Radiosonde data from weather balloons launched twice daily, providing temperature, dewpoint, and wind at multiple altitude levels
• Numerical weather prediction model output, the same models behind your TAFs
• PIREPs, ingested in something close to real time

The algorithm runs all of that through a physics-based set of equations designed to locate where conditions favor supercooled liquid water - droplets that exist in liquid form at temperatures well below freezing, as low as -40°C under the right conditions. The moment something solid disturbs them, they freeze on contact.

The conditions CIP specifically targets are temperatures between 0°C and -20°C, the presence of cloud or precipitation, and sufficient atmospheric moisture. Above -20°C, most water has converted to ice crystals, which don’t accrete on airframes the way liquid water does. That 0 to -20°C band is where classical structural icing lives.

What Does Icing Probability Actually Mean?

The output is a probability value at each point on a three-dimensional grid covering the continental United States. The grid updates hourly. When you see purple on an EFB icing overlay, you’re looking at a high probability zone. Lighter blues and greens indicate lower probability.

CIP also produces a separate severity estimate - trace, light, moderate, or severe.

Here’s what most pilots miss: this is a probabilistic product, not a deterministic one. A 70% icing probability doesn’t mean 70% of that airspace is full of ice. It means the algorithm assigns a 70% likelihood that icing conditions exist at that point at that time. A zone showing 20% probability is not safe - that’s still one in five.

Severity is harder to model than probability. Whether icing is trace or moderate depends on how much supercooled liquid water is present, the droplet size distribution, and how quickly it would accrete on a given airframe - which varies with aircraft speed, airfoil geometry, and surface temperature. Treat severity layers as guidance, not precision.

What Is the Forecast Icing Product (FIP)?

The Forecast Icing Product (FIP) runs the same algorithm forward in time using numerical model output, giving pilots a look at expected icing conditions up to 18 hours out. CIP is an analysis of current conditions built from real observations. FIP is what you’re using when you’re planning an afternoon flight over morning coffee.

Both products come from the Aviation Weather Center - NOAA’s operational arm for aviation weather - and flow to your EFB through data agreements.

Why Are Supercooled Large Droplets More Dangerous?

Standard structural icing is what the FAA calls Appendix C icing - small droplets that hit the leading edge of a wing and freeze there. When an aircraft is certificated for known icing conditions, that certification was historically tested against the Appendix C droplet size envelope.

Supercooled large droplets (SLD) are a different threat. These larger droplets carry more momentum, so they impinge farther back along the wing - past the leading edge where pneumatic boots or heated surfaces are effective. They can accrete ice on the empennage and downstream control surfaces where your protection system doesn’t reach.

The 1994 Roselawn accident was an SLD event. Ice accreted on the unprotected aft portion of the wing and disrupted airflow over the ailerons. The crew had no way to know they were in SLD conditions.

Post-accident, NCAR developed an SLD probability output to add to the icing product suite. That output exists today. Some EFBs display it; not all do by default. If you’re flying in conditions that could support SLD - particularly behind warm fronts in winter, with temperatures in the 0 to -12°C range where large drops survive - look for the SLD overlay specifically. The standard icing probability product does not fully capture that risk.

Why Are PIREPs an Imperfect Validation Source?

PIREPs are designed to be the real-world ground truth that validates and corrects the model. When the algorithm sees a pilot report of moderate icing at 10,000 feet over a certain location, it feeds back into the analysis and adjusts probability values in that area.

But PIREPs are inconsistent. They’re voluntary, so coverage is uneven. They’re subjective - a Cessna 172 and a Cirrus SR22 flying through the same air mass can have genuinely different icing experiences because of different speeds, airfoils, and accretion rates. And PIREPs are sparse over large parts of the country, particularly in the western United States, northern Alaska, and the central plains at night. In those regions, CIP leans more heavily on model output and satellite data, with less real-time pilot feedback to anchor it.

Some of this is being addressed through automated reporting. Major airlines participate in AMDAR (Aircraft Meteorological Data Relay), automatically downlinking temperature, wind, and icing encounter data multiple times per flight. Every major airline aircraft in cruise is effectively a flying weather station.

Whether that kind of automated reporting ever reaches the general aviation fleet meaningfully remains an open question. There are roughly 300,000 active GA aircraft in the United States, but the data density problem is significant - they’re not all flying all the time, and the airline network doesn’t serve the small airports where GA weather matters most. When automated GA meteorological reporting scales, icing products will improve substantially in those regions.

How Should You Actually Use CIP in the Cockpit?

Layer it with other sources. CIP tells you where conditions are probabilistically likely. PIREPs tell you what pilots have actually experienced. AIRMETs give you the forecaster’s judgment overlaid on the synoptic picture. Area forecasts and TAFs give you broader weather system context. No single layer answers the question; confidence comes from convergence.

Pay attention to the vertical gradient. Icing probability can change dramatically over a small vertical range. If heavy probability exists between 8,000 and 11,000 feet and tops are at 13,000, climbing to 14,000 may be a real option. Always know what’s above and below your cruise altitude, not just what’s at it.

Plan your exits before you enter the zone. Know where the nearest airport below the icing layer is and what the weather looks like there. Define your turn-back point before you’re accumulating ice faster than anticipated. That decision is much harder once you’re already in it.

Respect low probability zones. Twenty percent is one in five. Know your aircraft’s certification, understand the operational difference between anti-ice and de-ice, set your personal minimums, and treat the probability numbers honestly.

Key Takeaways

• CIP is a probabilistic product, not a guarantee. A 70% probability means the algorithm estimates a 70% chance icing conditions exist - not that 70% of that airspace is iced up.
• The critical icing temperature band is 0°C to -20°C, where supercooled liquid water persists. SLD events, typically in the 0 to -12°C range, can ice areas your certified protection system doesn’t cover.
• The 1994 Roselawn crash of American Eagle flight 4454 drove both regulatory change and NCAR’s development of an SLD probability output - which exists in CIP today but isn’t displayed by all EFBs by default.
• PIREPs are CIP’s real-world validation layer, but coverage is uneven - especially in the western U.S. and overnight. In data-sparse regions, the product relies more heavily on models and satellite data.
• FIP (Forecast Icing Product) extends the same algorithm up to 18 hours ahead and is the tool for preflight planning; CIP is the tool for current conditions analysis.