The Current Icing Product, the NCAR Sensor Fusion Algorithm That Turned Icing Forecasting Into a Three-Dimensional Probability Map, and What a Pilot Actually Needs to Understand Before Trusting the Colors

The Current Icing Product (CIP) is a real-time, three-dimensional probability map of atmospheric icing conditions, updated approximately every hour and freely available through the FAA’s Aviation Weather Center, ForeFlight, Garmin Pilot, and aviationweather.gov. Developed by the National Center for Atmospheric Research (NCAR) in Boulder, Colorado, it synthesizes data from six simultaneous sensor inputs through a probabilistic algorithm. The altitude slider in your preflight briefing is not a verdict - it is an estimate, and knowing how it was built determines whether you read it correctly.

What the Current Icing Product Actually Is

CIP is what engineers call a sensor fusion product. It does not rely on any single data source. The algorithm simultaneously ingests six major inputs and weights them to produce a probability value between 0 and 100 percent for each grid point in U.S. airspace, at every altitude from the surface up to 30,000 feet. Horizontal resolution is approximately 13 kilometers; vertical resolution is roughly 1,000-foot slices.

The FAA mandated its development. NCAR, working with aviation safety research funding, spent years in development and validation before the product reached operational status. CIP and its companion, the Forecast Icing Product (FIP), now live at the FAA’s Aviation Weather Center and feed into every major electronic flight bag.

The Six Data Sources Behind the Algorithm

Radar reflectivity. NEXRAD network data. Radar detects precipitation; precipitation in freezing air indicates probable icing.

Satellite imagery. Both visible and infrared, with particular attention to cloud top temperatures and cloud depth. Cold tops with significant vertical extent are a reliable marker for deep icing layers.

Numerical weather prediction model output. Temperature, relative humidity, wind, and pressure ingested through the full troposphere from National Centers for Environmental Prediction (NCEP) operational models.

Surface observations (METARs). Hourly airport reports that anchor the algorithm to observed ground-level conditions.

Pilot reports (PIREPs). The single most important input. When a pilot files an icing PIREP, that report enters the CIP algorithm as ground truth - real aircraft, real ice, real altitude. PIREPs are verifications, not inferences.

Lightning data. Flags convective cells where icing dynamics differ fundamentally from the stratiform icing that dominates most general aviation encounters.

What the Colors Actually Mean

The most common misread of CIP is treating it as a binary safe/avoid map. It is not. It is a probability map.

An orange region at your planned cruise altitude means atmospheric conditions are consistent with icing formation at a confidence level that demands attention - not that ice is guaranteed. A 60 percent probability does not mean you will find ice 60 percent of the time. It means that in historical cases with similar sensor signatures, icing was later verified approximately 60 percent of the time. Green means low probability. That is different from no ice.

FAA research into general aviation icing accidents found that a significant portion of pilots who encountered unexpected icing had completed a preflight briefing that included icing products. The failure was not checking. The failure was interpretation.

Why the Freezing Level Is Not the Icing Layer

These are two different things. Clouds can exist well below the freezing level and still contain supercooled water droplets - liquid water at temperatures below 32°F. In clean air, water can remain liquid down to approximately -40°F. In cloud, droplets can stay liquid well below freezing until an airframe provides the nucleation point. They freeze on contact. That is structural icing.

CIP accounts for this by incorporating cloud liquid water content, derived primarily from satellite data. An area that is cold but has very low liquid water content may show a lower icing probability than temperature alone would suggest. The combination of sub-freezing temperature, cloud presence, and high liquid water content is what drives the higher probability values.

The SLD Layer: What It Is and Why It Exists

CIP includes a dedicated Supercooled Large Droplet (SLD) layer that most general aviation pilots never access in a briefing. SLD are typically defined as droplets larger than 50 microns in diameter. Standard icing droplets freeze near the leading edge, where de-ice boots and heated surfaces are designed to handle them. Large droplets can flow back past the protected zone before freezing, accumulating ice in areas never evaluated during original certification testing.

This layer exists because of a specific accident. In October 1994, an American Eagle regional turboprop was holding at approximately 10,000 feet near Roselawn, Indiana, waiting to sequence into Chicago O’Hare. The aircraft entered an SLD environment. Ice accumulated in an unusual pattern that interacted with the autopilot and aileron control system. The aircraft rolled violently and impacted the ground. 68 people died.

The NTSB, in its investigation of American Eagle Flight 4184, cited the lack of SLD awareness and operational forecasting capability as a contributing factor. The FAA subsequently revised icing certification requirements and funded NCAR specifically to develop SLD detection and forecasting. The SLD layer in CIP is the direct technical response to Roselawn. If you fly in icing conditions and have never pulled that layer up during a briefing, find it now.

CIP vs. FIP: Which Product to Use When

CIP is an analysis product. It shows the best estimate of icing conditions in the past hour, based on assimilated sensor data. It is retrospective by nature.

FIP is a forecast product. It projects the same underlying physics forward using model output, in three-hour increments out to 18 hours from the current analysis time. For preflight planning, FIP is almost always the more relevant tool. You are not flying right now - you are flying in two hours, or five.

A Practical Preflight Workflow for Icing

For any IFR flight with icing potential:

Start with FIP at your planned altitude and departure time. Note probability values and severity estimates along the entire route.
Cross-check with CIP at the current analysis time. If observed conditions match what FIP predicted for this hour, the forward projection carries more weight.
Pull PIREPs from the last two to three hours along your route. A field of orange on FIP with zero icing PIREPs is a fundamentally different situation than the same field with four reports of moderate ice at your altitude in the last 90 minutes.

PIREPs are the ground truth. Everything else in the algorithm is modeled or remotely sensed.

Where CIP Can Mislead You

PIREP density. In high-traffic airspace during a busy travel day, the algorithm has substantial real-world verification feeding it. At 3 a.m. over rural Montana, it is estimating from satellite and radar with almost no real-aircraft confirmation. The product is least reliable when the sky is empty of reporting aircraft.

The update cycle. CIP refreshes approximately every hour. A developing squall line or afternoon convective buildup can produce a completely different icing environment than the product showed at your briefing. For rapidly evolving weather, treat CIP as a starting point, not a conclusion.

Mountainous terrain. Mountain wave environments create icing in configurations the algorithm does not model well. Wave clouds, rotor zones, and lenticular formations along ridgelines can produce severe icing in areas that appear benign on CIP. In the Rockies, Cascades, and Sierra Nevada, give CIP less weight and PIREPs more.

Aircraft capability is not encoded. CIP and FIP describe whether icing conditions exist in the atmosphere. They say nothing about whether your specific aircraft can handle what it finds. A King Air with full boots and heated surfaces can operate in conditions that would overwhelm a Cessna 172 with no ice protection. The product does not know what you are flying. That translation is the pilot’s job.

What’s Coming: Machine Learning Integration

NCAR’s Research Applications Laboratory has been working on integrating machine learning into the icing fusion algorithm. The traditional CIP approach uses a physics-based rule set: if temperature falls within a specific range, and liquid water content is at a given level, and radar reflectivity meets a defined threshold, then probability is X. A machine learning layer trained on decades of verified icing encounter data can identify non-linear relationships the rule-based system misses.

Early work presented at American Meteorological Society conferences showed meaningful improvement in detection rates, particularly for moderate and severe severity categories. That research is working through the transition from experimental to operational status.

Key Takeaways

CIP is a probability map, not a certainty map. Green means low probability, not ice-free. Orange means conditions consistent with icing formation, not guaranteed ice.
PIREPs are the most valuable input in any icing briefing. A high-probability zone with no recent icing PIREPs tells a different story than the same zone with multiple recent confirmations at your altitude.
FIP is almost always more relevant than CIP for preflight planning - you are analyzing a future flight, not current conditions.
Pull the SLD layer when icing is a factor. Supercooled large droplets accumulate ice behind de-ice boots, in areas original certification never evaluated. The Roselawn accident is why the layer exists.
CIP cannot account for your aircraft’s capability. The pilot must translate atmospheric probability into aircraft-specific risk - the product has no knowledge of what you are flying or what it is certified for.