Safety

How the US stopped airline wind shear crashes, and what's left

October 2, 202613 min readFor pilots and passengers

Leer en español · Lire en français · Ler em português

In this article

Four accidents that defined the problem
Fujita, NIMROD and JAWS: proving the microburst exists
Building the ground net: LLWAS, TDWR and WSP
In the airplane: reactive and predictive systems, and the training aid
What the data show
Wind shear has not gone away
The gaps in 2026
What it means for pilots

Between 1975 and 1994, four US airline crashes killed more than 400 people for the same basic reason: an airplane close to the ground flew into a microburst, the violent downdraft and outflow under a convective cell. The FAA's TDWR page names USAir 1016 at Charlotte, in July 1994, as the last wind shear related accident, and the scientists who built the warning systems wrote in 2022 that there had been no commercial microburst accident in the US since then (McCarthy et al., BAMS 2022). Both statements hold for airline flying, but not for every aircraft: in July 2022 a Part 135 Cessna 208B cargo flight crashed in a microburst at Salt Lake City, a TDWR airport (NTSB AIR-25-05).

That record came from research, radar, onboard sensors and training working together. It is worth understanding now because the hazard has not changed, the ground radars are more than 30 years old, and recent events in the UK, France and Salt Lake City show where the protection is thin.

Four accidents that defined the problem

Date	Flight	Aircraft	Phase	Killed
June 24, 1975	Eastern 66, New York JFK	Boeing 727-225	ILS approach, runway 22L	113 of 124 aboard
July 9, 1982	Pan Am 759, New Orleans	Boeing 727-235	Takeoff, runway 10	145 aboard, 8 on the ground
August 2, 1985	Flight 191, Dallas/Fort Worth	Lockheed L-1011	ILS approach, runway 17L	134 aboard, 1 on the ground
July 2, 1994	USAir 1016, Charlotte	DC-9-31	Approach and missed approach, runway 18R	37 passengers

Eastern 66. The 727 flew an approach through a very strong thunderstorm sitting on the localizer and hit the approach lights. The NTSB found the crew was slow to catch a high descent rate because they relied on visual cues, but added that the winds might have been too severe for a successful approach anyway. Minutes earlier, a Flying Tiger Line captain who had just landed told the tower about a "tremendous wind shear" on final and urged a runway change. The tower kept runway 22L in use (NTSB AAR-76-08).

Pan Am 759. The 727 lifted off, climbed to roughly 95 to 150 ft and descended into trees and houses. The NTSB found a decreasing-headwind shear of about 38 kt plus a downdraft at 100 ft, from a microburst on the airport. The Board listed the limited capability of ground-based detection as a contributing cause and recommended Doppler radar, microburst models in simulators and airborne detection (NTSB AAR-83-02).

Flight 191. The TriStar flew into the rain shaft beneath a storm with visible lightning on final. Among the probable causes, the NTSB cited "the lack of definitive, real-time windshear hazard information" and the lack of guidelines and training for avoiding and escaping low-altitude wind shear (NTSB AAR-86-05).

USAir 1016. The crew continued an approach into a storm over the airport. The captain called for a standard missed approach instead of a wind shear escape, thrust stayed below the go-around limit, and pitch fell to 5° nose down. The airplane's reactive wind shear system never alerted, which the NTSB traced to its software logic. Charlotte's TDWR had not been installed: a land dispute had pushed the site to 38th on the schedule. The NTSB also faulted the tower for not making sure controllers knew about, and reported, LLWAS alerts in multiple quadrants (NTSB AAR-95-03).

Fujita, NIMROD and JAWS: proving the microburst exists

University of Chicago scientist Ted Fujita studied the Eastern 66 data and proposed that a small, intense downdraft he called a "downburst" had caused it. He later named the smallest ones, 4 km across or less, "microbursts." Many scientists were skeptical (McCarthy et al.).

The proof came from Doppler radar. During the NIMROD project in northern Illinois in 1978, Fujita and NCAR's Jim Wilson saw a downburst on Doppler radar for the first time on May 29, and the campaign detected about 50 microbursts. The JAWS project near Denver's Stapleton Airport in 1982 detected 193. Work that followed JAWS showed that a microburst typically lives only 15 to 20 minutes and peaks within about 5 minutes of the downdraft reaching the ground, so any warning system had to update fast. The same paper notes the FAA did not fund JAWS until after Pan Am 759, which crashed during the field campaign.

Building the ground net: LLWAS, TDWR and WSP

LLWAS. The FAA says it developed the Low Level Windshear Alert System in 1977 in response to Eastern 66 (FAA WSDS). The first version compared a center-field anemometer with five others and flagged a 15-kt vector difference. It was built for gust fronts and sea breezes, and its sensors sat too far apart for microbursts. At the Flight 191 hearing, the JAWS director compared it to a net whose mesh is too coarse, so microbursts slip through (AAR-86-05). NCAR's expanded network version (LLWAS-NE) followed, and phase-3 systems went to 40 airports (McCarthy et al.).

TDWR. A prototype began in Boston in 1983, and the FAA formally started the acquisition in 1985 after Flight 191. Per the FAA, 45 TDWRs now protect 46 airports and have been in service since 1994 (FAA TDWR). The FAA had planned 105; funding issues held it to 45 (FAA WSDS). The AIM explains that TDWRs sit 8 to 12 miles from the airport and alert on microbursts and gust fronts in boxes ½ mile either side of each runway, out to 3 miles on final and 2 miles on departure.

ASR-9 WSP. The Weather Systems Processor adds wind shear detection to an existing ASR-9 surveillance radar. The FAA says it was developed in the 1990s in response to Flight 191 and is installed at 34 locations (FAA WSDS).

All three feed the same ribbon display in the tower, so the controller reads the alert as written, for example: runway 27 arrival, microburst alert, 35 kt loss, 2 mile final (AIM 7-1-24).

In the airplane: reactive and predictive systems, and the training aid

14 CFR 121.358, published in April 1990, requires turbine airplanes in Part 121 service to carry wind shear equipment. Airplanes built after January 2, 1991, and several listed models built before then, need an approved warning and flight guidance system, a detection and avoidance system, or both. Other older turbine airplanes need at least a warning system. Turboprops are excluded. The rule allows either type; it does not require forward-looking radar.

Reactive systems sense the shear once the airplane is in it, from airspeed, angle of attack, accelerations and pitch. Lincoln Laboratory notes the first was certified in November 1985 (Hallowell and Cho, 2010).
Predictive systems use the weather radar to measure the motion of rain in a downdraft ahead of the airplane. NASA flight tests gave 20 to 40 seconds of warning (AAR-95-03), and the first was certified in September 1994 (Hallowell and Cho).

Training was the third leg. In 1985 the FAA hired a team from Boeing, United, McDonnell Douglas, Lockheed California, Aviation Weather Associates and Helliwell, Inc. to produce the Windshear Training Aid, completed in 1987. Its Pilot Windshear Guide was reprinted for everyone as AC 00-54 in November 1988. The guide put avoidance first, then recognition, then a simple recovery. The FAA canceled AC 00-54 in December 2022 when it folded its weather ACs into one handbook.

What the data show

MIT Lincoln Laboratory's cost-benefit work for the FAA is the best public look at effectiveness (Hallowell and Cho, 2010; Cho and Hallowell, 2008):

From 1975 to 2006, the NTSB attributed 20 accidents and more than 500 deaths, across all aircraft categories, to wind shear. Accidents fell sharply even as traffic grew.
As of 2010, no wind shear accident had occurred at an airport with TDWR, WSP or upgraded LLWAS.
The authors credit pilot awareness and onboard predictive radar as well as ground systems, and estimate that pilot training alone cut wind shear safety exposure at the study airports, measured in cost terms, by about 26%.
A reactive system gives roughly a 37.5% (±12%) chance of recovering without ground contact. Predictive radar often exceeds 95% effectiveness in simulation, less in dry conditions.
Modeled detection: TDWR was the best single sensor for microbursts and gust fronts. WSP alone could not reach the required 90% microburst detection at many airports. A typical LLWAS detected about 50% of microbursts because its anemometers covered only part of the runway areas; Denver's full network reached 97%.

The limits of that evidence matter. Accidents were rare to begin with, and training, airborne systems and radars arrived together, so no one can assign the improvement precisely to any single fix.

Wind shear has not gone away

Recent reports show the hazard still bites, often where the 1990s systems do not reach.

Salt Lake City, 2022. A Part 135 Cessna 208B cargo flight hit a microburst during a go-around and crashed beside the runway; the pilot had minor injuries. The NTSB cited too few wind sensors and too little wind shear equipment in the tower as contributing. The tower's wind report came from one sensor 8,700 ft away while the ASOS showed gusts to 48 kt. The report also describes a 2016 Part 121 ERJ that saw airspeed jump from 190 to 234 kt on approach. In July 2025 the NTSB recommended LLWAS-NE for Salt Lake City and adequate wind detection at all susceptible Part 139 airports (NTSB AIR-25-05). Salt Lake City is on the NWS list of TDWR sites; the NTSB's findings centered on the tower's wind sensors.
Paris CDG, 2020. An Embraer 170 got a wind shear warning at 200 ft in a strong crosswind, flew the escape with wings level, and drifted toward an A320 departing the parallel runway. Separation reached 0.09 NM and 460 ft, with TCAS resolution advisories on both. The BEA found the procedures for those parallel operations had not adequately accounted for a wind shear escape in a strong crosswind (BEA2020-0471).
Prestwick and Edinburgh, October 3, 2025. During Storm Amy, a Boeing 737-800 flew three wind shear escape maneuvers on approaches at Prestwick and Edinburgh in gusty winds, then diverted to Manchester and landed with 363 kg of fuel, against a final reserve of 1,101 kg (AAIB 9H-QBD).

The gaps in 2026

Dry microbursts and gust fronts. The AIM warns that microbursts can come from benign-looking cells with virga, where a ring of blowing dust may be the only clue. Predictive radar needs moisture, and Honeywell says it cannot see a very dry microburst (Honeywell, 2019). Gust fronts can arrive many kilometers ahead of the storm. The AIM also lists what TDWR does not do: warn outside its alert boxes, detect shear that is not a microburst or gust front, or detect gusty crosswinds and turbulence.

Airports without radar coverage. Lincoln Laboratory's study covered 121 airports with ground-based systems plus 40 feeder airports with none. As of 2010, no predictive wind shear radar fit regional-jet or general aviation size limits (Hallowell and Cho).

The takeoff roll. Honeywell says predictive alerts are inhibited above 100 kt and below 50 ft, because continuing is judged safer at that point. The Pilot Windshear Guide warns that V1 may not be a valid go/no-go point in wind shear, since stopping depends largely on groundspeed, which can be much higher than airspeed (AC 00-54). LogAloft's founder has proposed, in a preprint under peer review, a groundspeed gate that restates V1 in groundspeed terms for this case. We cover it in depth in a separate article.

Lidar for clear air. Lidar measures wind from dust and aerosols and works best without rain, the opposite of radar. Hong Kong installed the first airport wind shear lidar in June 2002 and had four by 2024, covering 3 NM from each touchdown zone (HKO). Its lidar alerting system has run since the end of 2005, which the Observatory calls the world's first (HKO LIWAS). In the US, a lidar has been tested at Las Vegas alongside its TDWR since 2005 (McCarthy et al.), and Lincoln Laboratory modeled a radar plus lidar as the best combination.

Aging ground systems. In 2007, Lincoln Laboratory called removing TDWR in favor of available alternatives imprudent (Weber et al.). The FAA's February 2025 infrastructure roadmap shows TDWR Sustainment 3 running through the end of 2030, and a sustainment project for LLWAS and WSP through 2027 (FAA roadmap). The FAA says LLWAS and WSP have exceeded their planned 20-year service lives (FAA WSDS). It expects a future NextGen Surveillance and Weather Radar Capability to replace TDWR but gives no date (FAA TDWR). The January 2026 contracts to replace up to 612 surveillance radars by June 2028 do not mention TDWR. As of October 2026, we found no published TDWR replacement schedule.

What it means for pilots

Follow your company's procedures and your airplane's QRH. The points below come from the published guidance and the investigations above.

Avoid first. The Pilot Windshear Guide treats avoidance as the best defense, and avoidance can mean waiting. A microburst rarely lasts more than 15 minutes after it reaches the ground, keeps strengthening for about 5 minutes, and often comes with others nearby. Microbursts in a line can keep going for up to an hour (AIM 7-1-24).
Treat LLWAS and TDWR alerts as real. Eastern 66 and USAir 1016 both had warnings available that never changed the plan. A microburst alert with a 35 kt loss on final is a reason not to start the approach.
Know where coverage ends. TDWR boxes stop at 3 miles on final and 2 miles on departure. At an airport without TDWR or WSP, your own radar, PIREPs and the sky are most of what you have. Radar imagery can lag: the Salt Lake City pilot said the NEXRAD data he used en route was about 10 minutes old.
The escape, as published. For an encounter after liftoff or on approach, the Pilot Windshear Guide says: apply necessary thrust aggressively; pitch toward 15°, using intermittent stick shaker as the upper limit, and go higher if the flight path is still unacceptable; keep the flaps and gear where they are; do not trim for shear-induced speed changes; disregard flight director guidance not designed for wind shear. On approach, do not try to land. The pilot monitoring calls out sink rate and altitude (AC 00-54).
On the runway. The same guide says to reject before V1 only if airspeed variations are unacceptable and enough runway remains. After V1, continue, and rotate no later than 2,000 ft from the end even if VR has not been reached.
Report it. State the airspeed loss or gain and the altitude, not "negative wind shear." Tell ATC "wind shear escape" once the airplane is under control. Until you report the escape complete, ATC is not responsible for your separation (AIM 7-1-22). The CDG event shows why that matters next to a parallel runway.
Plan fuel for more than one escape. The Storm Amy crew used three escape maneuvers before running short.

Sources

This is general information, not operating guidance. Your company's procedures, your aircraft's manuals and the regulations come first.