#GPS#altitude#elevation#accuracy#geoid#WGS84

Why GPS Altitude Is So Inaccurate (and What the Numbers Actually Mean)

GPS vertical error is larger than horizontal error, and the height your device shows may be measured from an ellipsoid that sits tens of meters away from sea level. Here are the geometry, the geoid, worked Seoul and New York examples, and a practical checklist.

by 8 min readLast reviewed

Stand on a beach and a GPS receiver can report an altitude of −25 meters. Climb a summit marked 1,000 m and your watch may disagree by tens of meters. Horizontal GPS is usually good to a few meters, so why is the vertical number so much rougher? There are two separate reasons: satellite geometry and several different definitions of "zero."

I build drone ground control software, and in drone operations I have run into altitude readings that didn't match reality. Two things explain most such mismatches, and neither means the receiver is broken.

Reason 1: All the Satellites Are Above You

A receiver fixes its position by measuring ranges to satellites. Horizontally, those satellites are spread all around you (north, south, east and west), so errors in one direction are partly balanced by satellites on the other side. Vertically there is no such balance: every satellite you can use is above the horizon, and none is below you. Errors in the vertical axis are poorly constrained, and receivers express this as a larger vertical dilution of precision (VDOP) than horizontal (HDOP).

The official U.S. figures show it. The GPS Standard Positioning Service performance standard, as summarized on GPS.gov, specifies a global-average 95% positioning error of ≤ 8 m horizontal but ≤ 13 m vertical, and at the worst site ≤ 15 m horizontal versus ≤ 33 m vertical. Real devices vary with local conditions. GPS.gov's accuracy FAQ lists the main factors: satellite geometry, signal blockage, atmospheric conditions and receiver quality. It also notes that smartphones are typically accurate to about 4.9 m horizontally under open sky, which is a horizontal figure and doesn't apply to height.

Terrain and buildings make it worse. A valley or a street lined with tall buildings hides low-elevation satellites, and those are exactly the ones that help pin down height. Reflected signals (multipath) add their own error.

Reason 2: Which "Zero"? Ellipsoid, Geoid, or the Ground?

Even a perfect measurement can show a "wrong-looking" altitude, because there are three common references:

  • Ellipsoidal height (h): height above the WGS84 ellipsoid, the smooth mathematical shape GPS computes against. This is what the satellites natively give you.
  • Orthometric height (H): height above the geoid, the lumpy gravity-defined surface that mean sea level follows. This is the "elevation" on topographic maps and benchmarks.
  • Geoid height or undulation (N): the separation between the two at a given spot.

The U.S. National Geodetic Survey states the relationship plainly in its GPS on Benchmarks notes:

h = H + N        →        H = h − N

N is not small. In the EGM96 global geoid grid it ranges from about −107 m (in the Indian Ocean south of India) to about +85 m (near New Guinea).

Worked example: Seoul vs New York

I computed N at two cities with pyproj using the NGA EGM96 (15′) and EGM2008 (2.5′) geoid grids published on the PROJ data CDN:

LocationLat, LonN (EGM96)N (EGM2008)
Seoul City Hall37.5663, 126.9779+22.90 m+23.09 m
New York (Times Square)40.7580, −73.9855−32.64 m−32.62 m

The two models agree to within about 0.2 m at these points. The sign is the important part:

  • In Seoul the geoid sits about 23 m above the ellipsoid. A receiver that shows raw ellipsoidal height reads about 23 m too high compared with map elevation. For example, a hypothetical reading of h = 50.0 m corresponds to H ≈ 50.0 − 23.1 = 26.9 m above sea level.
  • In New York the geoid sits about 33 m below the ellipsoid, so raw ellipsoidal heights there read about 33 m too low. That explains the "negative altitude on the beach" effect.

Most phones apply a built-in geoid model before displaying altitude, but not every app, logger or drone field does. If two devices disagree by a roughly constant 20–30 m, check whether one shows h and the other shows H before assuming either is broken.

Barometric Altitude vs GNSS Altitude

Many watches, phones and nearly all drones also carry a barometer. It estimates height from air pressure, which falls as you climb.

  • Strengths: it is smooth and responds quickly to small changes in height. That makes it good for counting a staircase, logging cumulative ascent on a hike, or holding a drone level.
  • Weaknesses: air pressure also changes with the weather. A front moving through looks exactly like a slow climb or descent. Barometric altitude therefore drifts unless it is calibrated to a known elevation or blended with GNSS.

That's why good devices fuse the two: the barometer handles short-term changes and GNSS provides the long-term anchor.

AGL vs MSL vs Ellipsoid: The Drone Case

Drones add a fourth reference, the takeoff point. Many flight apps show altitude relative to where the aircraft took off. DJI's Mavic 3 Enterprise user manual, for example, describes the flight display's ALT as "the altitude of the aircraft relative to the takeoff point," shown separately from ASL, "relative to the average sea level."

These references can differ a lot on the same flight:

  • Relative to takeoff: 0 m on the launch pad, whatever the terrain.
  • AGL (above ground level): height above the ground directly beneath the aircraft, which changes as terrain rises or falls under you.
  • MSL / orthometric: height above the geoid, the number on aviation and topographic charts.
  • Ellipsoidal: what the raw GNSS solution and many mapping files use.

Take off from a hilltop and fly over a valley, and "50 m" relative to takeoff may be well over 50 m above the valley floor. Take off from a valley and fly toward a ridge, and the same reading can put the aircraft much closer to the ground than you think. When planning missions or reading logs, always confirm which altitude reference each field uses. Mixing them up is an easy way to misread altitude.

How RTK and PPK Fix the Vertical Problem

RTK (Real-Time Kinematic) and PPK (Post-Processed Kinematic) compare the rover's satellite measurements with those of a base station at a known position. Errors shared by both receivers, mainly satellite orbit/clock and atmospheric delay, cancel out, and carrier-phase measurements allow centimeter-level positions. GPS.gov notes that dual-frequency receivers and augmentation systems "can enable real-time positioning within a few centimeters."

The difference between the two, as Pix4D's RTK/PPK guide explains:

  • RTK applies corrections during the flight over a live link to a base station or network. Results are immediate, but the link must stay up.
  • PPK logs raw data on both receivers and applies corrections afterward. It is more resilient to signal interruptions, at the cost of a processing step.

When I've had drone position errors, switching to RTK or PPK is what resolved them. Two caveats apply to vertical work. First, the result is only as good as the base station's own coordinates. Second, RTK/PPK still produce ellipsoidal heights, so you must apply the correct geoid model to get elevations above sea level.

Practical Checklist

  1. Identify the reference. Is the number ellipsoidal (h), above sea level (H), above ground (AGL), or relative to takeoff? Look for labels such as "ellipsoid," "MSL," "ASL" or "ALT."
  2. Check the geoid offset for your area. A constant gap equal to the local N (about +23 m in Seoul, about −33 m in New York) means one device shows h and the other shows H.
  3. Give the receiver open sky. Wait for a stable fix away from tall buildings, cliffs and dense canopy.
  4. Don't trust a single reading. Average over a minute or more when you need a stable height.
  5. Calibrate the barometer at a known elevation and recalibrate when the weather changes.
  6. Use RTK/PPK or survey benchmarks when vertical accuracy really matters, such as for drainage, construction or mapping deliverables.
  7. Record the datum and geoid model (for example WGS84 + EGM2008) with any height you save or share.

FAQ

Why does my phone show a different altitude than the map? Usually it's a mix of vertical GNSS error, which is larger than horizontal error, and sometimes a reference mismatch. If the gap stays roughly constant, compare it with the local geoid undulation.

Is barometric altitude more accurate than GPS altitude? For short-term changes in height, usually yes, because it is smoother. For absolute height it drifts with the weather unless calibrated, so the best devices combine both.

Does RTK give me height above sea level? Not directly. RTK and PPK give very precise ellipsoidal heights. You still need a geoid model to convert them to orthometric (sea-level) heights.

The Takeaway

Horizontal coordinates and altitude come from the same satellites, but they aren't equally reliable. Geometry favors the horizontal, and the vertical answer also depends on which zero you mean. To see your current fix, open My Location. For a broader look at what degrades GPS readings, such as the atmosphere, multipath and geometry, see our GPS accuracy guide. The fundamentals of the measurement itself are covered in how GPS works.

Your latitude and longitude remain the dependable part. You can view your current fix in every format with the GPS coordinates finder, or convert readings with the coordinate converter.

Related Posts

#GPS#accuracy#satellite

How Accurate Is GPS? Real Numbers: 5 m on Phones, 1 cm with RTK

How accurate is GPS? Typically about 5 meters on smartphones — down to centimeters with RTK. See accuracy by device type and the factors that degrade it, with practical tips.

13 min read
Read More
#GPS#GLONASS#Galileo

GPS vs GLONASS vs Galileo vs BeiDou: The Four Global Navigation Systems Compared

GPS is just one of four global satellite navigation systems. Compare GPS, GLONASS, Galileo, and BeiDou (orbits, satellite counts, civil signals), plus QZSS, NavIC, and correction methods like SBAS, RTK, PPK, and PPP.

9 min read
Read More
#GPS#satellite#trilateration

How GPS Actually Works: Satellites, Signals, and Trilateration

A clear explanation of how GPS determines your position: the satellite constellation, radio signals traveling at light speed, atomic clocks, relativity corrections, and why you need at least four satellites.

8 min read
Read More