Terrain_awareness_and_warning_system

# Terrain awareness and warning system > [[PORTAL_Aviation|Aviation]] · [[PORTAL_Avionics|Avionics]] spine.  ## Microsims — three.js ### Terrain Awareness and Warning System (three.js) <div class="microsim-player"> <iframe src="https://wikitube-3d-microsims.netlify.app/flight/Terrain_awareness_and_warning_system.html" width="100%" height="620" frameborder="0" loading="lazy" sandbox="allow-scripts allow-same-origin" title="Terrain Awareness and Warning System — three.js microsim"></iframe> </div> **Open it full-screen:** [flight/Terrain_awareness_and_warning_system.html](https://wikitube-3d-microsims.netlify.app/flight/Terrain_awareness_and_warning_system.html) · library `threejs` · route `microsim/threejs/` ### Related microsims Live sims on neighbouring articles: - [[Controlled_flight_into_terrain]] - [[Traffic_collision_avoidance_system]] - [[Radio_navigation]] - [[Head-up_display]] - [[Autopilot]] - [[Avionics]] *Sim hosted off-article; the article owns the reference, not the runtime (WIKI_RULES §10.4).*  ## Overview A terrain awareness and warning system (TAWS) is the airborne equipment that tells a flight crew, out loud and without being asked, that the aeroplane is about to fly into the ground. It exists because of one unglamorous accident type: **controlled flight into terrain**, in which a serviceable aircraft under the full control of a qualified crew is flown into ground, water or an obstacle. Nothing breaks; the crew simply do not know where the ground is. For the first four decades of jet transport this was the largest killer in commercial aviation, and the two generations of equipment described here are among the clearest safety wins in the record. The first generation was the **ground proximity warning system** (GPWS), developed in the late 1960s and early 1970s largely by C. Donald Bateman and his team at Sundstrand Data Control. It worked from what was already on the aeroplane: a radio (radar) altimeter, air data, the glideslope receiver and the gear and flap position switches. Out of those it built a family of alerting laws numbered Modes 1 to 5, later extended to 6 and 7 — excessive descent rate, excessive rate of closure with terrain, altitude loss after take-off, unsafe terrain clearance for the configuration, excessive deviation below the glideslope, advisory callouts and windshear. After a run of fatal CFIT accidents, the FAA made TSO-approved GPWS mandatory for large turbine-powered aeroplanes in US air carrier service in 1974 (14 CFR 121.360), phased in over the following year. ICAO followed with an Annex 6 standard applying from 1 July 1979 to aeroplanes over 15,000 kg or certificated for more than 30 passengers. The rule change was answered by history almost immediately. On 1 December 1974, TWA Flight 514, a Boeing 727 descending toward Dulles in poor weather, struck the western slope of Mount Weather, Virginia, killing all 92 people on board. It had no GPWS. GPWS worked, and the airline CFIT rate fell sharply through the late 1970s and 1980s. But it had a structural limitation that no amount of tuning could remove. **A radio altimeter looks straight down.** It measures the distance to the surface underneath the aircraft, and nothing else. GPWS therefore had to infer danger ahead from the behaviour of a number that describes only what is below. Mode 2 does this by differentiating the radio altitude: if the ground below is rushing up at you, terrain is presumably rising. Against gently rising ground that inference works. Against a cliff, a mesa edge or a steep ridge face it fails, because until the aeroplane actually crosses the foot of the obstacle the ground below is flat and the closure rate is zero. By the time the number moves, the rock is in the windscreen. The accident that made this argument unanswerable was **American Airlines Flight 965**, a Boeing 757 that struck El Deluvio ridge near Cali, Colombia, on 20 December 1995 after a navigation error took it into a valley parallel to the intended one. The GPWS did fire, and the crew responded immediately and correctly — autopilot out, full power, nose up to about 30 degrees. The FAA's own case study records that the system gave **only 13 seconds of warning before impact**, and that the speedbrakes, extended during the descent, were never retracted during the escape. 159 people died. The investigation concluded that more advance notice of a terrain conflict was needed if ground-proximity warnings were to be survivable. The second generation answers that by changing the question. An **enhanced GPWS** (EGPWS, the Honeywell trade name; TAWS is the generic and regulatory term) carries a stored digital elevation model of the world's terrain, takes its own position from GPS or the flight management system, projects the aeroplane's flight path forward for roughly a minute, and tests whether any stored terrain rises into that predicted corridor. It keeps all the old radio-altimeter modes as a backstop, and adds a terrain display that paints the surrounding country in a colour scheme keyed to the aircraft's own altitude. Caution alerts are issued typically 60 seconds ahead of a predicted conflict and warnings typically 30 seconds ahead — an order of magnitude more time than Mode 2 can offer against steep ground. The regulatory response was rapid. The FAA issued its TAWS final rule on **29 March 2000** (65 FR 16736), creating 14 CFR 91.223, 121.354 and 135.154 and requiring TAWS on all US-registered turbine-powered aeroplanes configured with six or more passenger seats. Newly manufactured aeroplanes had to comply from 29 March 2002; the existing fleet had until **29 March 2005**. Equipment is approved against TSO-C151: Class A (all the GPWS modes plus the forward-looking function and a mandatory terrain situational-awareness display) is required for Part 121 and for Part 135 aircraft with ten or more seats; Class B (the safety functions without a required display) covers Part 91 turbine aircraft with six or more seats and smaller Part 135 operations. ICAO's Annex 6 now requires a ground proximity warning system **with a forward-looking terrain avoidance function** for turbine aeroplanes above 5,700 kg maximum certificated take-off mass or authorised to carry more than nine passengers. Europe carries an equivalent requirement in Commission Regulation (EU) No 965/2012, Annex IV (Part-CAT). The result is one of the least ambiguous safety improvements in the record. ICAO reports that the CFIT accident rate has fallen by roughly a factor of seven since third-generation aircraft with ground-proximity equipment entered service, with close to 99 per cent of flights now operated with terrain warning systems fitted. CFIT has not vanished — IATA counted 47 CFIT accidents worldwide between 2008 and 2017, 42 of them fatal, causing 892 deaths — but it now runs at roughly 0.05 accidents per million sectors for jets against 0.47 for turboprops, and the fleet in which it still happens is disproportionately the one least likely to be equipped. Two caveats must be stated plainly, and the simulation on this page is built to make both of them visible. **The database can be wrong, and the position fix can be wrong.** IATA's own risk assessment of EGPWS database validity notes that operators update terrain and obstacle databases on manufacturer release cycles rather than on any risk-based schedule, that out-of-date data can both suppress legitimate alerts and generate false ones, and that repeated false alerts corrode the crew habit of responding to real ones. Honeywell's pilot guides say directly that the terrain display "does not provide the accuracy and/or fidelity to be the sole source for deciding terrain or obstacle avoidance", and that the forward-looking functions **automatically inhibit themselves** if no position source meeting their accuracy requirement is available. Nor is the equipment sufficient by itself: IATA's review of 51 CFIT-related events found that in about 47 per cent of them the crew did not respond adequately to the alert that was given. ## The physics ### What a radio altimeter actually measures A radar altimeter transmits downward through a broad-beam antenna — total beamwidths of roughly 40 to 60 degrees are typical — and locks onto the **shortest slant range** returned from inside that beam. Over flat ground that is the vertical distance. Over sloping or broken ground it is not: the receiver finds the nearest piece of dirt anywhere in the cone. If the aircraft is at altitude *z* above mean sea level and the surface is *h(x, y)*, the reported radio altitude is ``` RA = min over the footprint of sqrt( (z - h(x', y'))^2 + r^2 ) ``` where *r* is the horizontal offset of the sample point from the aircraft. The simulation models this with a 50-degree beam and samples the centre plus two rings, which is why the modelled Mode 2 gets a second or two of grace at a cliff — the edge of the cone clips the rock face slightly before the aeroplane is over it. That is a real effect, and it is not nearly enough. ### Mode 1: excessive descent rate Mode 1 compares barometric descent rate against radio altitude. Below a boundary that rises with sink rate the system says "SINKRATE"; below a second, tighter boundary it says "PULL UP". The published envelope appears in the pilot guides only as a graph; the simulation uses a two-point linear fit whose endpoints are stated in the source (caution from 1,500 ft/min at 50 ft AGL to 5,000 ft/min at 2,450 ft AGL; warning from 1,900 ft/min to 7,000 ft/min over the same span). That fit reproduces the familiar teaching values of roughly 2,900 ft/min for SINKRATE and 3,900 ft/min for PULL UP at 1,000 ft AGL. Mode 1 needs no terrain knowledge at all and is genuinely good at what it was built for — a descent that will not stop over ground that is not rising. ### Mode 2: excessive terrain closure Mode 2 differentiates radio altitude. Write the terrain height along track as *h(x)*, the aircraft altitude as *z(t)*, the ground speed as *V* and the vertical speed as *ż*. Then ``` RA(t) = z(t) - h(x(t)) d(RA)/dt = z_dot - V * dh/dx closure = -d(RA)/dt = |sink rate| + V * (terrain slope) ``` The whole behaviour of the reactive system falls out of that second term. Over flat ground `dh/dx = 0` and the closure rate is just the descent rate. Over a slope rising at 2,000 ft per nautical mile at 280 kt (4.67 NM/min) the terrain contributes 9,300 ft/min of closure and Mode 2 fires hard. Over a **vertical** face `dh/dx` is effectively a spike: the closure rate is zero right up to the foot of the cliff and then unbounded, and the alert arrives with no useful time attached to it. Mode 2 also has a ceiling. The published upper radio-altitude limit is 1,650 ft, expanding linearly to 2,450 ft as airspeed increases from 220 kt to 310 kt. Above that limit the mode is simply **not armed**, because at high radio altitude any plausible closure rate is a normal descent. A jet cruising 2,600 ft above a plateau is therefore outside the Mode 2 envelope altogether — a fact the cliff scenario in the simulation exploits, and one that is entirely faithful to the equipment. When the terrain display is enabled the manufacturer *reduces* the Mode 2A ceiling (to 1,250 ft, and to 950 ft in later software), precisely because the forward-looking function is now doing the work at longer range. ### The look-ahead test The predictive function replaces differentiation with lookup. Let *x*<sub>gps</sub> be the position the navigation system believes the aeroplane occupies, *T* the look-ahead time, *C* the required terrain clearance, and *h*<sub>db</sub> the stored database elevation. The system declares a conflict when, for any τ between zero and *T*, ``` h_db( x_gps + V*tau ) > z + z_dot*tau - C ``` evaluated across a search ribbon that starts a quarter of a nautical mile wide and spreads ±3 degrees laterally, taking the highest terrain across its width. The right-hand side is the **floor** of the protection volume: the predicted flight path, dropped by the clearance requirement. The simulation draws that volume as an actual swept box — an amber caution wedge of length *V·T* with a red warning wedge of length *V·T*/2 nested inside it, matching the guide's 60-second caution and 30-second warning. Terrain that penetrates the volume is painted solid on the surface, which is exactly what the real display does when it paints a "terrain threat area". Three properties of this test matter. **Climb credit is limited.** A crew who are already climbing should not be able to talk the system out of an alert by pointing the nose up; the projection caps the upward credit it will grant. The model uses 500 ft/min. **The clearance floor is not constant in the real equipment.** En route it is several hundred feet; near a runway it must shrink to a couple of hundred, or every normal landing would trigger it. Honeywell implements this as a separate Terrain Clearance Floor function keyed to distance from the nearest runway centre, with a minimum around 245 ft adjacent to the runway. This simulation uses a single en-route value of 700 ft and no airports, which makes its flat-terrain scenario alert earlier than a real installation would. **The database is a grid of maxima, not a surface.** Terrain data are stored as cells — 0.5 NM is close to the classic 30 arc-second en-route cell, with much finer grids around airports — and each cell records the **highest** point inside it. That conservatism is deliberate: a database that rounded peaks down would kill people. The visible consequence is that a sharp escarpment appears to the system as a step half a cell early, and that ridge crests generate alerts slightly sooner than the smooth ground would justify. The simulation draws these cells as cyan plates floating above the true surface so the staircase is visible. ### Why position error is a timing error Along-track position error translates directly into time. If the fix is wrong by *e* nautical miles, every look-ahead sample is displaced by *e*, so the alert moves by ``` delta_t = e / V ``` At 280 kt, one nautical mile is 12.9 seconds. Two nautical miles of lag costs 26 seconds — which, against a 60-second caution, is most of the margin. In the simulation the whole protection volume detaches from the aeroplane and slides along the track, tethered to a cyan "GPS believes here" marker, and the numbers move accordingly: on the cliff scenario, 2 NM of lag drops the TAWS caution lead from 64 s to 39 s and the subsequent escape manoeuvre no longer clears the wall. A lead error is not harmless either; it produces nuisance alerts over ground that is not a threat, and nuisance alerts are how crews learn to ignore the box. ### Why lead time, not altitude, is the currency The published escape is: disconnect the autopilot, apply maximum thrust, roll wings level and rotate toward the pitch limit or stick shaker. Model the vertical response as a recognition delay *t*<sub>r</sub> followed by a vertical acceleration *a*. Arresting an existing descent of *v*<sub>0</sub> costs ``` altitude lost = v_0 * t_r + v_0^2 / (2a) ``` which for a 2,000 ft/min descent, a two-second delay and 0.25 g is about 67 ft of delay plus 69 ft of arrest — around 135 ft. That is almost nothing. The binding constraint is not altitude, it is **gradient**. A transport jet at low level sustains perhaps 2,500 ft/min of climb, which at 280 kt is 535 ft per nautical mile, a flight path angle near 5 degrees. Terrain does not care: the ridge in the mountain scenario rises at up to 2,170 ft per nautical mile, about 18 degrees. Once the terrain gradient exceeds the achievable climb gradient, no amount of pulling will out-climb it — the only thing that can save the aeroplane is having started the manoeuvre far enough back that the terrain has not yet risen above the reachable altitude. Distance in hand is the only currency, and lead time is how you buy it. Running the four scenarios in the simulation with default settings gives the comparison in simulated seconds of warning before the un-escaped trajectory reaches the ground: | Terrain profile | Reactive GPWS lead | Predictive TAWS lead | Difference | |---|---|---|---| | flat plain, 2,500 ft/min descent | 17.7 s (Mode 1) | 76.8 s | +59 s | | rolling hills to 2,100 ft | 15.9 s (Mode 2) | 72.7 s | +57 s | | 2,700 ft escarpment | **4.4 s** (Mode 2) | 64.5 s | +60 s | | 9,000 ft ridge (Cali geometry) | 11.7 s (Mode 2) | 63.9 s | +52 s | The mountain figure is worth pausing on. A radio-altimeter system flying a jet at 280 kt into a ridge rising across the track gets about a dozen seconds. The Cali crew got thirteen. ## Controls -> what each maps to | Control | Symbol | Range and units | What it changes | |---|---|---|---| | Terrain profile | *h(x, z)* | flat / rolling / cliff / mountain | Selects the synthetic elevation model and resets the scenario to an initial condition that makes that profile's lesson visible inside about a minute. Rebuilds both the rendered surface and the gridded database. | | Terrain database | — | ON / OFF | ON annunciates the predictive look-ahead and draws the protection volume (Class A TAWS behaviour); OFF leaves only the radio-altimeter Modes 1 and 2 (1974-generation GPWS). Both philosophies are always *computed*, so the lead-time comparison is live either way; the switch decides which one is allowed to speak and to trigger the escape. | | Descent rate | *ż* | 0 to 4,000 ft/min down | Commanded vertical speed. Feeds Mode 1 directly, feeds Mode 2 through the closure-rate term, and sets the slope of the projected flight path that forms the roof of the look-ahead volume. | | Airspeed | *V* | 120 to 340 kt | True airspeed, taken equal to ground speed (no wind is modelled). Sets the look-ahead **range** *V·T*, scales the terrain-slope contribution to Mode 2 closure, and expands the Mode 2A ceiling between 220 and 310 kt. | | Look-ahead time | *T* | 15 to 90 s caution | Length of the caution volume in time; the warning volume is always *T*/2, matching the published 60 s / 30 s pair. Longer *T* buys earlier alerts and more nuisance exposure. | | GPS along-track error | *e* | −3 to +3 NM | Displacement between the true position and the position the navigation solution reports. Negative means the fix lags the aeroplane and the alert comes late; positive means it leads and the alert comes early. Costs or gains *e*/*V* seconds. | | Vertical exaggeration | *k* | 1 to 12 × | Drawing only. At *k* = 1 the geometry is honest and nearly unreadable: a 2,700 ft escarpment is 0.44 NM tall inside a 4.7 NM look-ahead. Every readout stays in true feet. | | Camera | — | chase / profile / plan | Chase looks where the aeroplane is going; profile shows the vertical geometry of floor against terrain; plan reads like a navigation display. | | Time rate | — | 0.5 to 4 × real time | Wall-clock compression. All reported times are simulated seconds, so the comparison is unaffected. | | Fly the escape on warning | — | on / off | When on, the aeroplane flies the modelled escape (2 s recognition, 0.25 g slew, 2,500 ft/min target climb) on the *active* system's warning. The counterfactual shadow trajectory never escapes, which is what keeps the lead-time reference honest. | | Protection volume / beam / database cells | — | on / off | Show or hide the swept caution and warning wedges, the radio-altimeter cone, and the gridded database plates. | | Pull up now (P) | — | button / key | Manual escape, ignoring the alerting logic. | | Pause (Space), Step (S), Reset (R) | — | — | Freeze, advance two simulated seconds, or restart the scenario. Sliders remain live while paused, so the look-ahead scan repaints as you move them. | ## Learning objective After using this simulation you should be able to say, without hedging, **why** a forward-looking terrain system is not merely a nicer version of a ground-proximity system. Specifically: that a radio altimeter measures a scalar describing the ground below and can only infer the ground ahead by differentiating it; that this inference degrades exactly in proportion to how steep the terrain is, so it is weakest against the terrain most likely to kill you; that a database plus a position fix plus a projected flight path replaces the inference with a direct query; that the resulting alert arrives roughly a minute rather than roughly ten seconds before impact; and that the extra time matters because an aeroplane cannot out-climb a mountain — it can only start earlier. You should also be able to state the two failure paths of the new philosophy from having watched them: an out-of-date or wrong database, and a position solution that is not where the aeroplane is. ## Limits and connections This is a teaching model, and it simplifies aggressively. **Flight mechanics.** The aeroplane flies a fixed heading in still air with no bank, turn, wind or performance model, and speed is commanded rather than computed. The escape is kinematic — a fixed recognition delay, a fixed vertical-acceleration slew and a fixed target climb rate, with no thrust, drag, altitude or weight dependence, and no modelling of the effect that mattered at Cali: extended speedbrakes silently eating the climb performance the crew thought they had. **Alerting.** Only Modes 1 and 2 of the reactive family are implemented. Modes 3 (altitude loss after take-off), 4 (unsafe terrain clearance for the configuration), 5 (glideslope), 6 (advisory callouts and bank angle) and 7 (windshear) are absent, as are the Terrain Clearance Floor and Runway Field Clearance Floor functions and the obstacle database. There are no airports in the model, so the single en-route clearance floor of 700 ft is never reduced the way a real installation reduces it on approach; that makes the flat-terrain scenario alert earlier than reality. The Mode 1 and Mode 2 envelopes are **fits** to the shapes and stated endpoints published in the pilot guides; the exact curves are manufacturer data and are not published numerically anywhere the author could check. Any specific threshold quoted from this simulation should be treated as illustrative of the shape of the envelope, not as an equipment specification. **Position and altitude.** Only along-track position error is modelled. Real position error is two-dimensional and would also move the search ribbon laterally, which against a valley wall is arguably worse. Vertical error is not modelled at all, and it is a genuine issue: the look-ahead test compares a database elevation in feet above mean sea level with the aeroplane's own altitude, and barometric altitude is not a geometric quantity — it depends on the altimeter setting and on the temperature profile, and reads *low* in cold air. Honeywell addresses this with a computed "Geometric Altitude" blending GPS altitude, barometric altitude, runway elevation and terrain data. None of that machinery is here. Neither is the real system's self-inhibit: a real EGPWS switches its look-ahead functions off when it cannot establish position accuracy, whereas this model happily keeps computing against a fix you have deliberately corrupted. **Database.** The database in this model is generated from the same function that draws the terrain, so it is never *wrong* — only coarse and conservative. Real terrain databases have contained real errors, and real obstacle databases are incomplete outside well-surveyed regions; IATA has raised database currency as a standing safety issue, noting that update cycles are driven by manufacturer release schedules rather than by risk. Modelling a genuinely erroneous cell would be a straightforward extension and an instructive one. **Attribution.** It is easy, and probably correct, to credit TAWS with a large share of the collapse in the CFIT rate; it is much harder to isolate that share. The same decades brought crew resource management, stabilised-approach criteria, GNSS approaches with vertical guidance replacing dive-and-drive non-precision procedures, better charting, and the Flight Safety Foundation's CFIT and approach-and-landing reduction campaigns. Terrain warning is one large factor among several correlated ones, and no published analysis known to the author cleanly separates them. [[Controlled_flight_into_terrain]] is the accident category the equipment exists to prevent. [[Traffic_collision_avoidance_system]] solves the structurally identical problem for other aircraft rather than for the ground, and solves it with cooperative interrogation instead of a stored map — the contrast between a database you carry and a transponder that answers you is a deep one. [[Radio_navigation]] covers the position sources TAWS depends on, and whose failure at Cali created the accident in the first place. [[Head-up_display]] and [[Autopilot]] are the other places where the aeroplane's own idea of the world is shown to, or acted on by, something other than the pilot's eyes; [[Avionics]] is the parent article for the family. ## References 1. Honeywell International Inc. *MK V and MK VII Enhanced Ground Proximity Warning System (EGPWS) Pilot's Guide*. Document 060-4241-000, Rev. D, March 2000. 2. Honeywell International Inc. *MK VI and MK VIII Enhanced Ground Proximity Warning System (EGPWS) Pilot's Guide*. Document 060-4314-000, Rev. C, May 2004. 3. Federal Aviation Administration. "Terrain Awareness and Warning System." Final Rule, 14 CFR Parts 91, 121 and 135. *Federal Register*, Vol. 65, No. 61, 29 March 2000, p. 16736 (docket 29312, amendment 00-7595). Creates 14 CFR 91.223, 121.354 and 135.154. 4. Federal Aviation Administration. *Technical Standard Order TSO-C151, Terrain Awareness and Warning System*. Washington, DC. (Classes A, B and C.) 5. Federal Aviation Administration. *14 CFR 121.360, Ground Proximity Warning-Glide Slope Deviation Alerting System*. (The 1974 GPWS requirement.) 6. International Civil Aviation Organization. *Annex 6 to the Convention on International Civil Aviation — Operation of Aircraft, Part I: International Commercial Air Transport — Aeroplanes*. Montreal. (Ground proximity warning system with forward-looking terrain avoidance function; thresholds of 5,700 kg maximum certificated take-off mass and nine passengers.) 7. European Union. *Commission Regulation (EU) No 965/2012 of 5 October 2012 laying down technical requirements and administrative procedures related to air operations*, Annex IV (Part-CAT). Official Journal of the European Union L 296. 8. Aeronáutica Civil of the Republic of Colombia. *Aircraft Accident Report: Controlled Flight Into Terrain, American Airlines Flight 965, Boeing 757-223, N651AA, near Cali, Colombia, December 20, 1995*. Santafé de Bogotá, 1996. 9. Federal Aviation Administration. *Lessons Learned From Transport Airplane Accidents: American Airlines Flight 965, Boeing 757-223, N651AA*. FAA Lessons Learned library. (Source for the 13-second GPWS warning interval and the un-retracted speedbrakes.) 10. National Transportation Safety Board. *Aircraft Accident Report: Trans World Airlines, Inc., Boeing 727-231, N54328, Berryville, Virginia, December 1, 1974*. Report NTSB-AAR-75-16, Washington, DC, 1975. 11. International Air Transport Association. *Controlled Flight Into Terrain Accident Analysis Report, 2008–2017 Data*. IATA, Montreal, 2018. (47 CFIT accidents, 42 fatal, 892 fatalities; crew-response breakdown across 51 GPWS/TAWS-related events.) 12. International Air Transport Association. *EGPWS Database Safety Risk Assessment*. IATA Safety Issue Hub, external publication. (Database currency as a standing safety issue; false and suppressed alerts; erosion of crew trust.) 13. Flight Safety Foundation. "Killers in Aviation: FSF Task Force Presents Facts About Approach-and-landing and Controlled-flight-into-terrain Accidents." *Flight Safety Digest*, Vol. 17 No. 11–12 / Vol. 18 No. 1–2, November 1998 – February 1999. 14. Commercial Aviation Safety Team. *Safety Enhancement SE001: Terrain Awareness Warning System (TAWS) — Final Report*. CAST, 24 May 2006. 15. RTCA, Inc. *DO-161A, Minimum Performance Standards — Airborne Ground Proximity Warning Equipment*. Washington, DC, 1976. 16. National Business Aviation Association. *Terrain Awareness and Warning Systems (TAWS) and Ground Proximity Warning Systems (GPWS)*. NBAA operations resource. (Summary of the 1974 GPWS mandate, the 2000 rule and the 29 March 2005 retrofit deadline; Class A and Class B applicability by operating rule.) 17. International Civil Aviation Organization. *Aviation Safety: Ground Proximity Warning Systems*. ICAO Uniting Aviation. (1979 Annex 6 thresholds; the sevenfold reduction in CFIT rate; CFIT as roughly a quarter of fatalities from about 3 per cent of accidents.) 18. United States Patent and Trademark Office and the National Science and Technology Medals Foundation. C. Donald Bateman, recipient of the US National Medal of Technology and Innovation for 2011, for the invention and development of ground proximity warning and terrain awareness systems. **On the spine:** [[Avionics]] · [[Fly-by-wire]] · [[Autopilot]] · [[Head-up_display]] · [[Traffic_collision_avoidance_system]] · [[Terrain_awareness_and_warning_system]] · [[Radio_navigation]] · [[Radiation_hardening]] · [[Radar]] · [[Air_traffic_control]] · [[Aviation]].  **Part of the [[Avionics]] hub** — main article for section X19, *Terrain awareness*. Related sections: [[Traffic_collision_avoidance_system]] · Global Positioning System · Aviation accidents and incidents.  ## Wikipedia : Wikitube **Strict pair:** [Wikipedia](https://en.wikipedia.org/wiki/Terrain_awareness_and_warning_system) : [Wikitube](https://en.wikitube.io/wiki/Terrain_awareness_and_warning_system) --- *PORTAL_Avionics three.js batch · 2026-08-05 · sim staged in `_3d_deploy_stage/`.*