The Current Icing Product, the Multi-Sensor Fusion Algorithm That Solved Aviation's Three-Dimensional Icing Blind Spot, and the Free Tool That Predicts Structural Ice by Altitude Layer Every Hour

The Current Icing Product (CIP) and its companion, the Forecast Icing Product (FIP), replace vague regional AIRMET polygons with a three-dimensional, hourly map of structural icing probability at specific altitude layers. Developed by the National Center for Atmospheric Research (NCAR) in partnership with NOAA and the FAA, CIP first went operational in 2002. Both products are free at aviationweather.gov and represent the most precise icing guidance available to any pilot planning flight into clouds.

Why Wasn’t the AIRMET Enough?

The AIRMET Sierra is still part of a complete weather brief - but its fundamental limitation is resolution. A forecaster draws a polygon covering several states and declares icing “possible” below a certain altitude. That call is categorical: yes or no, based on model output and satellite data. The actual high-probability icing layer might be a thin band from 6,000 to 9,000 feet while the AIRMET reads “below 12,000 feet.”

The fatal gap in that system became undeniable on October 31, 1994, when American Eagle Flight 4184, an ATR-72, lost control while holding in a cloud deck over Roselawn, Indiana. 68 people died. The crew was operating in an icing environment that appeared manageable, but those clouds were saturated with supercooled large droplets (SLD) - droplets large enough to flow behind the leading-edge deice boots and freeze on surfaces the system was never designed to protect. The forecasting tools of the time could not characterize that threat at the altitude and location where the aircraft was holding.

Roselawn drove a fundamental shift in how the aviation weather community approached icing information. “Possible in this region” was no longer an adequate answer. Pilots needed to know what kind of icing, at what altitude, and how severe. The research program that followed fed directly into CIP’s development.

How Does the CIP Algorithm Actually Work?

CIP is a physics-based algorithm that ingests multiple data streams simultaneously, every hour, to produce a three-dimensional icing probability field. Five inputs drive the fusion:

1. GOES satellite data. The Geostationary Operational Environmental Satellite system provides continuous cloud property coverage across the continent. Cloud-top temperature is a critical variable - cold tops indicate deep systems with high liquid water content. Modern satellite processing can also distinguish between clouds dominated by ice crystals versus those with significant supercooled liquid water, which is the substance that actually accretes on airframes.

2. NEXRAD radar. The Next Generation Radar network feeds reflectivity data into the algorithm. High reflectivity in a cold-cloud environment signals heavy ice water content. The dual-polarization upgrade completed around 2013 made this input substantially more useful - by transmitting and receiving energy in both horizontal and vertical planes simultaneously, the radar can now distinguish between rain, snow, hail, and mixed-phase precipitation in ways older single-polarization systems could not.

3. METARs. Surface observations from thousands of automated reporting stations establish the boundary conditions: temperature, dew point, current weather, cloud layers. The algorithm knows what’s happening at the ground before it reasons about what’s happening above.

4. The Rapid Refresh model. This is the structural backbone. NOAA’s Rapid Refresh is an operational mesoscale model that runs every hour, ingesting surface observations, radar, and satellite data to produce a three-dimensional atmospheric analysis - temperature, relative humidity, cloud water content - at every pressure level across a grid covering North America.

5. PIREPs. Pilot reports are ingested in near-real time as calibration data. If the algorithm predicts low icing probability in a layer and three PIREPs are reporting moderate ice there, the product adjusts. Real-world observation takes explicit priority over model output when the two conflict.

The output is a gridded, three-dimensional probability field. Icing probability is mapped at altitude increments from the surface to 30,000 feet, in 2,000-foot steps. At a specific grid square and altitude, the product might read: 62% probability of structural icing encounter in the next hour.

What Is the FIP and When Should You Use It?

The Forecast Icing Product takes the same three-dimensional output framework and projects it forward using Rapid Refresh forecast fields - out to 6, 12, and 18 hours. For preflight planning, the FIP is often more operationally useful than the current analysis, because the CIP tells you what the atmosphere looks like right now while the FIP tells you what it is likely to look like when you are actually airborne.

The most effective way to use both products is to compare them. Pull the current CIP first, then the FIP for your departure time. The difference between the two reveals the trend: is the high-probability band growing or shrinking? Is the freezing level shifting? Is a clean window developing? The FIP does not guarantee accuracy, but trend information against the current analysis is genuine intelligence for route planning.

Both products are free at aviationweather.gov under the Icing section. The interactive map lets you select altitude layers, overlay probability shading, and step through forecast hours.

How Do You Build CIP Into Your Preflight Workflow?

Work the altitude dimension. This is the most powerful and most underused capability of the product. Don’t look at one altitude level and make a decision - step through every level along your route. The icing threat may be concentrated in a narrow band you can fly above, fly below, or transit quickly on the climb. The altitude resolution is exactly what separates CIP from a text AIRMET, and ignoring it means using a three-dimensional tool in two dimensions.

Cross-check CIP against the AIRMET and any applicable SIGMET. If the AIRMET covers a broad area but CIP concentrates the high-probability zone in one specific corner, that’s signal about where the real threat lives. If CIP and the AIRMET both converge on the same area and altitude, that agreement is strong confirmation. Independent products that agree increase confidence; products that diverge demand explanation before you commit to the route.

Check the SLD layer product separately. The supercooled large droplet layer product uses the same data fusion framework as CIP and flags areas where droplet size distribution suggests the large-droplet threat. Supercooled large droplets have enough mass to travel past protected leading-edge zones and freeze on unprotected surfaces behind deice boots - regions the certification assumptions never covered. SLD indicated on your route at your altitude is a get-out-of-that-layer signal, regardless of your deice equipment.

File PIREPs every time you are in clouds at icing temperatures. Report ice if you find it. Report nothing on the airframe if that is what you find - negative icing PIREPs are calibration data just as valuable as positive ones. The system is explicitly designed to use real-world feedback to correct model output. The report you file tonight directly helps the next pilot on the same route.

Where Does CIP Fall Short?

Complex terrain is the biggest limitation. The NEXRAD network has coverage gaps in mountainous regions because radar energy travels in a straight line and mountains block it. The Rapid Refresh model’s horizontal resolution does not fully capture small-scale lifting and orographic enhancement on windward slopes. A dry-looking synoptic pattern over the Rockies can still produce significant localized icing that CIP misses entirely. In mountain flying, treat CIP as one input among several, not the definitive answer.

CIP can lag in rapidly evolving situations. The product updates every hour. If a sharp frontal boundary moves faster than the model expected, the analysis can be behind reality by enough to matter. Look at the trend - is the threat area growing or shrinking? Is the freezing level rising or falling? The evolution tells you as much as the current state.

Probability values require interpretation against your specific aircraft. A 60% icing probability means conditions are significantly favorable for an icing encounter - not that you will definitely hit ice, and not that 30% means you are safe. A Cessna 172 in a 60% probability zone is a fundamentally different situation than a Piper Navajo Chieftain with deice boots in the same zone. CIP does not know your airplane. That judgment belongs to you.

Where Is Icing Forecasting Technology Headed?

CIP was the product of a specific convergence: the computing power to run hourly mesoscale models, NEXRAD reaching national coverage, the GOES satellite series reaching operational maturity, and the regulatory pressure following the 1994 Roselawn accident. Each of those conditions had to be in place simultaneously.

The next generation of improvement is coming from automated aircraft observations. The Tropospheric Airborne Meteorological Data Reporting (TAMDAR) system installs environmental sensors on regional jets that automatically report temperature, humidity, wind, icing encounters, and turbulence in real time at flight levels. These observations feed directly into models like the Rapid Refresh and are far more numerous and systematic than the voluntary PIREP network. As the TAMDAR sensor fleet expands, the input data quality for CIP and FIP continues to improve.

Aviation weather information today is better than at any point in the history of the IFR system. The research behind CIP and FIP is published by NCAR and publicly available. The Aviation Weather Center provides product methodology documentation worth fifteen minutes of any instrument pilot’s time.

Key Takeaways

• CIP maps icing probability in three dimensions at 2,000-foot altitude increments up to 30,000 feet, updated every hour - a fundamentally different product from a text AIRMET polygon.
• Five data streams drive the algorithm: GOES satellite, NEXRAD radar (dual-polarization since approximately 2013), METARs, the Rapid Refresh mesoscale model, and near-real-time PIREPs - with real-world pilot reports taking priority over model output when they conflict.
• Use CIP and FIP together: CIP shows current conditions, FIP projects the trend for your departure time. Comparing the two reveals whether the threat is building or clearing.
• The SLD layer product is a separate check - supercooled large droplets defeat deice systems designed for standard icing, and SLD indicated on your route is a hard avoidance signal.
• CIP has real limitations in complex terrain, in fast-moving frontal situations, and in accounting for your specific aircraft’s capability. Use it as one layer in a complete brief, not as a standalone clearance.