Wx Watch: News From the Icing Front

It's been more than three years since the icing-related crash of American Eagle Flight 4148, an ATR-72 that fell from the sky in Roselawn, Indiana. That crash will go down in aviation history as a turning point in icing education, forecasting, and rulemaking. It's also one more affirmation that bureaucracies require major loss of life to shake them from their torpor and begin moving with a purpose.

To briefly recap, the ATR crashed after a loss of control was caused by flying in supercooled-large-droplet (SLD) icing conditions of the kind previously unacknowledged by the aviation community. Researchers had documented this kind of SLD icing before, but their findings received little heed or publicity outside of professional circles.

American Eagle 4184 was holding at 10,000 feet in SLD icing, then cleared to descend to 8,000 feet. In an autopilot turn, one wing stalled and the airplane rolled over into a tight spiral, hitting the ground at an estimated 375 knots and killing all 68 persons aboard (see " Wx Watch: The Worst Ice," December 1995 Pilot).

The NTSB determined that the roll excursion was probably caused by an aileron hinge moment reversal — in other words, an uncommanded movement of an aileron. That uncommanded movement was probably caused by a rearward shift in the wing's center of lift, which was caused by a buildup of ice that had run back from the wing leading edge area protected by the airplane's pneumatic deice boots. It was the large size of the water droplets that allowed the supercooled droplets to run this far back on the wings before freezing. Droplet size was estimated at between 100 and 2,000 microns in diameter. Those are big droplets; a typical raindrop is 1,000 microns in diameter.

The airplane wasn't designed to handle this type of ice. It turns out that no airplane is. The icing certification envelopes require manufacturers to prove their ice protection systems in droplet diameter conditions — 15 to 50 microns — much smaller than those found in SLD conditions. Forecasters didn't know how to predict SLD icing because it was unknown in FAA circles.

Other gross shortcomings also came out of the NTSB's Roselawn crash investigation: At the time, HIWAS and other weather reports and forecasts didn't give icing intensities or precise altitudes affected by icing; ATR knew about roll anomalies when the ATR-72 is flown in ice, but didn't reveal them, the NTSB asserted (a separate report, issued by the French equivalent of our FAA, disputed this claim, and a war of words ensued); and icing forecasts aren't specific enough or updated quickly enough. That's the gist of many, many findings the NTSB made. As for recommendations, the NTSB asked, in part, that the FAA come up with more accurate icing "nowcasts" with two-hour valid times, and update the aircraft icing certification envelopes to reflect the new knowledge about SLD conditions. It also asked for pilot education programs about SLD, for the FAA to make sure that the FARs are compatible with the published definition of severe icing, and for the elimination of implied authorization of flight into severe icing conditions for aircraft certified for flight in those conditions. As we should all know by now, a manufacturer's certification of an airplane for flight in known icing doesn't give you carte blanche to fly in the worst conditions.

NTSB recommendations, however, are just that. They don't carry regulatory force. The best the NTSB can hope for is that the FAA (or whatever other agency, company, or organization it addresses) will turn the recommendations into official rules or guidelines.

In April 1997 the FAA published its "In-Flight Aircraft Icing Plan," an ambitious effort aimed at modernizing our understanding and operational procedures with regard to icing — especially of the supercooled-large-droplet kind. The plan is divided into 14 different task categories among five working groups: icing environmental characterization; ice protection and ice detection; forecasting and avoidance; requirements for and means of compliance in icing conditions; and operational regulations and training requirements.

The forecasting community is perhaps the first to act on the recommendations directed their way. Even predating the Roselawn crash, meteorologists at the National Center for Atmospheric Research (NCAR) in Boulder, Colorado, have been examining new ways of presenting more accurate and timely icing forecasts. Since the crash, the pace and scope of this research have increased to truly admirable levels.

One new forecasting tool is what's being called an "observation-based stovepipe algorithm." It's being used in computer models to graphically depict areas, and altitudes within these areas, suspected of having all types of icing conditions — including the SLD variety. This algorithm works by first searching for surface observations of freezing drizzle, freezing rain, or ice pellets. Then it searches the gridded fields of a more conventional forecast computer model for any areas of temperature or relative humidity conducive to icing, as revealed by research of the conditions existing when pireps of the worst icing were reported. If surface observations don't show any freezing rain, drizzle, or ice pellets, then the algorithm simply searches for any type of precipitation or overcast skies. (The stovepipe moniker refers to the vertical column of air that the algorithm searches.)

There's also a forecast-based stovepipe algorithm. It's a blend of surface precipitation predictions from two other computer models: the Rapid Update Cycle (RUC), and the ETA model.

"The stovepipe algorithms are far better than the 'blind' algorithms we used before," said Marcia Politovich, an NCAR meteorologist who specializes in winter weather and who is an adviser to the Icing Plan's steering committee. "Those just found temperatures and humidities. At least with the stovepipes we find an overcast to go with the right conditions aloft.

"As for the forecast-based model, we had the same problem, with the result that the icing areas were overpredicted. Now that seems to be fixed and ... the verification of the predictions with actual icing has turned out better than we expected.

"As a matter of fact, we're beating current airmets in terms of the size of airmet areas. So the overwarning problem is much less. Having satellite input also helps us greatly in predicting where clouds will be — which is a far bigger problem than you might think," Politovich added.

New algorithms are also being tried out at the Aviation Weather Center (AWC, the originator of advisories such as sigmets and airmets), where satellite-derived cloud-top temperatures, model input, Nexrad radar plots, and surface observations are also combined in still another computer model.

This all may sound very esoteric, but the practical benefits of these new models and algorithms will be striking once they're made operational. We've all heard dire forecasts of icing conditions that advertise entire regions loaded with scary ice, and canceled our flights accordingly. Or, worse, some pilots — who may have tired of forecasters crying wolf every time a cloud deck appears — launch anyway, in the belief that any ice won't be as bad as forecast.

If only we could really trust an icing forecast or had a way to know how to circumnavigate the areas really affected by ice, we keep telling ourselves. The stovepipe and other new computer-derived products of the future may soon be to icing what Nexrad and Terminal Doppler Weather Radar are now to thunderstorms and wind shear.

Safer circumnavigation of icing conditions could come as early as this month, when the AWC is scheduled to begin using the graphics products produced by the new methods described here.

You don't have to be tucked away in a cubicle at NCAR or the AWC to see this new model output, by the way. You can have a very close look at what forecasters will use by summoning the following NCAR Web page over the Internet ( www.rap.ucar.edu/largedrop/).

In far greater detail than would a magazine article, this page will tell you just what NCAR has been up to over the past few years. You can call up the bases and tops of icing conditions, as well as scan several layers of the atmosphere for all types of icing conditions. The photograph accompanying this article is a scan of one of the stovepipe pages and shows the bases of icing conditions.

Icing pireps can also be called up, and so can icing airmets and sigmets. As if this weren't enough, there are links to all manner of other standard weather reports, forecasts, and AWC products, too. There's even a photo of SLD icing on the leading edge of a NASA Lewis Research Center Twin Otter; you can see the ice running back, well aft of the Otter's deice boots.

This new information can't come soon enough. Too bad it took a highly publicized crash to bump up its implementation. In an upcoming "Wx Watch," we'll take a look at some of the regulatory initiatives that came from the Icing Plan. With a total of 18 icing-related notices of proposed rulemaking, the implications are far-ranging and would affect the flight rules governing many of the FAR Part 23 airplanes that we fly most often.

Links to all Web sites referenced in this issue can be found on AOPA Online. E-mail the author at [email protected].