The first sign is almost always the same. The magenta line on the navigation display does not vanish dramatically; it simply stops agreeing with itself. A position jumps a few miles sideways. The autopilot shrugs. Somewhere in the stack, a box decides it no longer trusts what it is being told, and a message appears that most crews now recognise on sight.

Over the Baltic, the eastern Mediterranean, the Black Sea and increasingly over the western United States, this has stopped being a curiosity. GPS interference is a routine operational condition. Which raises a question that sounds naive until you try to answer it properly: if the satellites go quiet, what is actually left?

The answer is a surprising amount. Aviation spent its first seventy years solving this problem without any satellites at all, and most of what it built is still bolted to the airframe, still transmitting from a field in Germany, or still spinning quietly in a box behind the cockpit. This is what a crew falls back on, in the order they fall back on it.

Pilotage: the method that never stopped working

Strip away every box in the aircraft and one technique survives: look out of the window and compare what you see with a chart. Pilotage is the first thing a student pilot learns and the last thing that can be jammed, because the river, the ridge line and the motorway junction are not transmitting anything.

Its limits are obvious and worth stating plainly. Pilotage needs daylight, decent visibility and terrain with features worth recognising. It is close to useless over an ocean, over desert, over a uniform forest, or above a solid undercast. It does not scale to a jet covering eight miles a minute. But at low level, in weather, with everything else failing, it is the floor beneath the whole structure, and military low-level training still treats it that way.

The modern version is slightly less romantic and considerably more useful: a crew cross-checking a terrain display and a chart against what the windscreen is showing, and noticing when the three stop agreeing.

The clearest explanation on the internet of how an inertial platform knows where it is without being told anything. Worth the time before reading the inertial section below.

Dead reckoning and the wind triangle

Dead reckoning is the arithmetic of not being lost. You know where you started. You know your heading, your airspeed, how long you have been flying and what the wind is doing. From those, you can compute where you must now be.

The complication is the wind, and the wind is the whole discipline. An aircraft flies through a moving block of air, so the direction it points and the direction it travels are different things. Resolving the two is the wind triangle: aircraft heading and true airspeed on one side, wind direction and speed on another, and the resulting track and groundspeed on the third.

For most of a century that triangle was solved on a circular slide rule. The E6B and its descendants put the whole problem on two sides of a plastic disc, and it is still carried in flight bags and still examined on checkrides, for the straightforward reason that it needs no electricity and cannot be spoofed.

Dead reckoning degrades predictably rather than suddenly, which is its great virtue. Every error compounds with time, so a position becomes steadily less certain rather than abruptly wrong. A crew that knows its last confirmed fix and its elapsed time since can draw an honest circle of uncertainty around itself, and that circle is a far more useful thing to hand to air traffic control than a confident number from a box that has been lying.

Radio navigation: the NDB, the VOR, and the network America is deliberately keeping

The first great leap away from eyeballs and arithmetic was the ground-based radio beacon. The non-directional beacon is the simplest of them: a transmitter sitting on the ground radiating equally in all directions, and an automatic direction finder in the aircraft whose needle points at it. Fly so the needle stays on the nose and you will arrive, eventually, having been pushed sideways by every crosswind on the way.

The VHF omnidirectional range was the serious answer. A VOR broadcasts a reference signal and a rotating signal, and the phase difference between them tells the receiver which of 360 radials the aircraft is sitting on. Add distance measuring equipment and the aircraft knows not just the bearing but the range, which is a position rather than a direction. Those beacons, in their distinctive cone-and-counterpoise housings, built the airway system.

The interesting part is what happened next. Satellite navigation made most of these beacons redundant, and the obvious move was to switch them all off. The FAA decided not to. Instead it built the Minimum Operational Network, a deliberately thinned but deliberately preserved skeleton of VORs kept running specifically so that aircraft have somewhere to go when GPS does not work.

The numbers are specific. The FAA's discontinuance programme covers 302 VORs, of which roughly 200 have been switched off so far, with the remainder scheduled through to the 2030 fiscal year. What survives is 594 stations. The network is designed to give near-continuous VOR signal coverage at and above 5,000 feet above ground level, and to leave every designated MON airport within 100 nautical miles of at least one conventional instrument approach, whether an ILS, a localiser or a VOR approach.

That is not neglect. That is a national aviation authority spending money to keep obsolete infrastructure alive because it has decided a single point of failure is unacceptable. In the mountainous west the retained service volume was actually widened, from 40 nautical miles at 5,000 feet in the 2015 network to 70, precisely because terrain and distance make the fallback harder there.

Distance measuring equipment got the same treatment. The FAA's NextGen DME programme is building out coverage to support DME/DME/IRU area navigation in 62 major terminal areas, which lets suitably equipped aircraft fly modern procedures from ground stations alone. Note the IRU in that string. Even the ground-based backup leans on an inertial reference unit to hold things together between fixes.

LORAN: the chain that was switched off, and keeps being argued back

Long Range Navigation was hyperbolic navigation, and the idea behind it is elegant. A chain of ground stations transmits pulses in a precisely timed sequence. A receiver measures the difference in arrival times between a master station and its secondaries. Each time difference places the receiver somewhere on a hyperbolic curve; two curves intersect; that intersection is your position.

LORAN-C worked at 100 kHz, low enough to follow the curvature of the earth and reach far out over water, which is exactly where celestial and pilotage were weakest. It covered oceans. It worked at night, in cloud, in weather that made a sextant useless.

The Smithsonian National Air and Space Museum on how hyperbolic navigation actually works, with the time-difference geometry drawn out.

It was killed by economics rather than by failure. In May 2009 the US administration declared the system obsolete; a Department of Homeland Security appropriations measure signed on 28 October 2009 cleared the legal path to terminate it. The Coast Guard gave notice in January 2010 and began decommissioning on 8 February, with transmissions ending that year.

The condition attached to that law is the part worth remembering. Termination required a certification that the infrastructure was not needed as a backup to GPS. The pilot community argued at the time that this was exactly backwards, and the subsequent decade of jamming and spoofing incidents has not made that argument weaker. Enhanced LORAN has been repeatedly authorised as an option for resilient positioning and timing without ever quite being built, and the towers that were not demolished are still, periodically, the subject of serious proposals.

There is a reason the idea refuses to die. A LORAN signal is a ground-based transmission at enormous power and very low frequency. Jamming it is a different and much harder proposition than overpowering a satellite signal that arrives at the antenna weaker than background noise.

Inertial navigation: the box that remembers every movement you made

Everything above depends on something outside the aircraft. A chart needs ground you can see. A VOR needs a transmitter. LORAN needs a chain. Inertial navigation needs nothing at all, and that is the whole point of it.

An inertial system is told once, on the ramp, exactly where it is. From that moment it measures its own acceleration in three axes and its own rotation about three axes, and integrates. Acceleration integrated once gives velocity; integrated twice gives distance. Keep track of which way the sensors are pointing and you can carry a position forward indefinitely without receiving a single photon from the outside world. It cannot be jammed, because it is not listening. It cannot be spoofed, because there is nothing to spoof.

The price is drift. Every sensor has a small bias, and integration is merciless with small biases: an error in acceleration becomes a growing error in velocity, which becomes an accelerating error in position. The published arithmetic is sobering. A horizontal gyro drifting at one hundredth of a degree per hour produces roughly 0.7 nautical miles of position error after a single hour. An azimuth gyro at two tenths of a degree per hour gives about a mile in the same time.

Good hardware fights this to a draw. A high-grade strapdown ring-laser system is specified at around 0.6 nautical miles per hour in pure inertial mode, with platform gyros holding bias stability in the region of a few thousandths of a degree per hour. That is the figure that made transoceanic flight without a navigator possible, and it is why a modern airliner losing GPS mid-Atlantic does not become lost. It becomes slightly less certain, at a known and predictable rate.

One elegant detail runs underneath all of it. An inertial platform is tuned to the Schuler period of about 84 minutes, the period of a hypothetical pendulum whose length equals the radius of the earth. Tune the system that way and errors in the vertical reference oscillate around the truth over 84-minute cycles instead of growing without limit. It is a piece of early twentieth-century physics quietly holding up twenty-first-century aviation.

Curious Droid on how inertial navigation reshaped air, sea and space travel, including the ballistic-missile work that paid for it.

This is also where the military answer diverges from the civil one. If drift is the only enemy, the solution is to correct it occasionally from something unjammable, and the oldest unjammable reference of all is the sky. The Spirito B-2 carries an astro-inertial system that does exactly this, taking star sights in daylight to trim its inertial solution, a technique we covered separately and which is the direct descendant of what follows.

Celestial: the port in the ceiling

Before inertial platforms were small enough or cheap enough, long-range flight carried a specialist whose entire job was the sky. The navigator's instrument was the bubble sextant, which solved the problem that makes marine celestial navigation impossible in an aircraft: there is no reliable horizon. Instead of measuring a star against the sea, a bubble sextant measures it against an artificial horizon formed by a bubble in a liquid, the same principle as a spirit level.

Taking the sight meant getting an unobstructed view of the sky, which is why so many wartime aircraft grew a transparent blister in the roof. The C-47 Skytrain, the Liberator, the Short Stirling and the Avro Lancaster all carried astrodomes, and the technique survived into the early jet bombers, the Canberra and the V-force among them.

Astrodomes had two problems. They added drag, and in a pressurised aircraft they were a structural weakness that could fail explosively, occasionally killing the navigator using it. The fix was the periscopic sextant, inserted through a small sealed port rather than a dome, and the port is the detail that outlived everything else.

Schiff's account is worth reading for the detail that the port was still there after the people who used it had gone, and that on his first Atlantic crossing a flight engineer opened it by accident. It was loud. It barely touched the cabin pressure. By then, as he puts it, the last of the airline's professional celestial navigators had been retired.

Celestial navigation is not coming back to the airlines. But it is the proof of concept for every unjammable navigation idea now being funded, from astro-inertial on bombers to the magnetic-anomaly and vision-based systems currently in flight test.

What actually happens today

Put it together and the real answer to "how do pilots navigate without GPS" is that they do not fall back to one thing. They fall back through a stack.

The inertial reference system keeps flying the aeroplane, drifting slowly and honestly. DME/DME updating trims that drift wherever ground stations are dense enough, which is most of the developed world's terminal airspace. The VOR network, preserved on purpose, provides the conventional airways and approaches when area navigation is not trustworthy. Radar vectors from a controller who can still see the aircraft are often the simplest answer of all. And underneath everything sits dead reckoning and a crew that knows the time, the heading and the wind.

The equipment that does this is not glamorous. Some of it is older than the crew. But the reason a GPS outage over the Baltic is an annoyance rather than an emergency is that aviation never quite got around to throwing any of it away.

Sources: Federal Aviation Administration, VOR Minimum Operational Network and NextGen DME programme briefing (2025); GPS.gov, LORAN-C Infrastructure and E-LORAN policy record; AOPA, “Coast Guard sounds death knell for loran-C” (January 2010); Barry Schiff, Proficient Pilot, AOPA Pilot (May 2010); iMAR Navigation, “Inertial Navigation: Forty Years of Evolution”; Smithsonian National Air and Space Museum; Wikimedia Commons.