Table of Contents
A runner finishes a loop through a city park, opens the activity map, and finds a straight line cutting across a pond, over a fence, and along the roof of a bus shelter. The distance reads about 300 m short of the figure on the trailhead sign. Nothing is wrong with the watch. The trace is what happens when a receiver has to choose between satellites it hears clearly and satellites it can barely hear at all.
GPS accuracy is not one number and it is not fixed. It changes street by street and minute by minute, and it changes most in exactly the places people like to run: under trees and between tall buildings. This article explains what degrades the signal, how to read a trace that looks wrong, and which habits actually improve the recording.
The Four Variables Behind Every Position Fix
Every position your watch records is a calculation, not a reading. The receiver listens for timing signals from satellites, works out how far away each one is, and solves for a point that fits all of those distances. Four things shape how well that solution holds up:
- Satellites in view. A clear sky gives a receiver 10 to 14 usable satellites in most of the world. Dense canopy can cut that to five or six.
- Geometry. Satellites spread across the sky triangulate well. Satellites bunched in one direction leave the position loosely constrained, which is what dilution of precision, or DOP, describes.
- Signal strength. A weak signal carries more noise, and noise turns into meters of error.
- Multipath. Signals that bounce off a surface arrive late and make the satellite look farther away than it is.
The watch reports a single latitude and longitude, but the confidence behind that point can vary by an order of magnitude from one block to the next.
Why Tree Cover Is Harder Than Cloud
Clouds are mostly transparent to the radio frequencies GPS uses. Leaves are not. Foliage holds water, and water absorbs and scatters the signal, so a wet summer canopy can knock several decibels off every satellite before the signal reaches your wrist. A single layer of broadleaf canopy might cost 1 to 3 dB. A dense evergreen stand with wet needles can cost much more.
The effect is not a dead signal but a thinner one. The receiver loses the weakest satellites first, usually the ones low on the horizon that would have improved geometry. What remains sits more directly overhead and is more clustered, so the position starts to wander. Under heavy cover the watch may also fall back to smoothing: it holds the last good fix and extrapolates, which produces the telltale straight segment through a bend.
Deciduous trees in leaf are worse than the same trees in winter, and rain makes both worse again.
Multipath: The Bounce That Moves You Sideways
Multipath is the reason city traces look unsteady. A satellite signal can travel directly to the watch or bounce off a glass facade, a metal roof, a parked bus, or a stretch of water first. The reflected path is longer, so the receiver computes a larger distance to that satellite and shifts the solution away from the truth.
Errors from multipath are usually tens of meters, not hundreds, but they change quickly. One fix lands 15 m east, the next 20 m west, and the map draws a jagged line that inflates distance and makes pace look erratic. Glass towers and waterfront promenades are classic offenders because smooth surfaces reflect cleanly.
You cannot outrun multipath, but you can avoid feeding it. Running down the center of a wide street instead of hugging a building line gives the receiver a better view of the sky and fewer reflective surfaces close by.
Urban Canyons and the Geometry Problem
A street of tall buildings is a canyon: a narrow strip of sky, walls on both sides. The receiver can still see satellites almost directly overhead, but it loses the low-elevation ones that would pin the position down horizontally. With every visible satellite in a similar part of the sky, the solution becomes strong vertically and weak along the street.
That asymmetry produces a specific symptom. The trace follows your progress along the block reasonably well but jumps sideways by 10 to 30 m, sometimes crossing to the other pavement or into a building footprint. Distance may be nearly correct while the shape is wrong.
DOP captures this. Values below about 2 are good, 2 to 5 is workable, and above 5 means the geometry is poor. Consumer watches rarely display it, but some activity exports include it, and a spike in DOP lines up neatly with the ugly part of the map.
What Dual-Frequency GPS Actually Changes
Most watches receive the L1 signal at 1575.42 MHz. Dual-frequency models also receive L5 at 1176.45 MHz. Because the ionosphere delays the two frequencies by different amounts, a receiver that hears both can estimate and remove most of that delay. L5 also carries a higher chipping rate, which helps separate a direct signal from a reflected one.
The gain is real but situational. In open terrain, single and dual frequency often agree within a few meters. In a canyon or under wet canopy, dual-frequency tracks tend to hold their shape better and recover faster after a bad stretch. The improvement shows up as fewer wild spikes rather than a magically perfect line.
Dual-frequency reception costs battery, because the receiver has more work to do. On a long outing, check whether the mode is set to all systems or a power-saving combination, and decide which matters more: a clean trace or two extra hours. The trade-offs are similar to those covered in battery strategies for long GPS watches.
Why the Route Map Cuts Corners
A watch does not record a continuous line. It records a position every second, or every few seconds in smart-recording mode, and the map draws straight segments between those points. On a switchback or a tight corner, the straight segment cuts across the inside of the turn.
At running speed this is small: a 90-degree corner taken at 3 m/s loses roughly a meter of path. On a hairpin descent or a winding trail, the losses add up, and a route can measure 1 to 3 percent short even with excellent reception. Add smoothing, which some watches apply to make the line look tidy, and the innermost corners get trimmed further.
This is why two watches on the same wrist can disagree about a twisty 10 km loop while agreeing almost exactly on a straight out-and-back. The disagreement lives in the corners, not in the satellites.
Reading Distance Errors: Short, Long, or Both
Distance errors run in two directions, and the direction tells you something.
- Too long usually means noise. Scattered fixes around your true path add zigzag length. A trace that looks fuzzy, with many small side-to-side spikes, will read high.
- Too short usually means interpolation and dropouts. Straight segments across bends, or a stretch where the watch lost lock and rejoined later, remove distance.
- Both at once is common in cities: spikes inflate one section and a canyon cuts another.
A reasonable expectation for a consumer watch in mixed terrain is within 1 to 3 percent of a surveyed course. If you routinely see more than 5 percent, the problem is worth fixing, and it is usually fit, firmware, or a recording setting rather than the satellites alone. The troubleshooting steps in fixing inaccurate GPS tracking start with those same variables.
Habits That Improve the Recording
Most of the fix happens before you press start.
- Give the watch a clear view for 30 to 60 seconds before starting. A cold start needs to download the satellite almanac, and moving immediately pushes that work into the first kilometer.
- Start in the open, not under a porch or a tree. The first fix sets the tone for the opening minutes.
- Wear the watch snug and on the outside of the wrist when possible, so it does not slide over the antenna.
- Use the highest recording rate for races and interval sessions, and smart recording for long easy days where battery matters more.
- Do not stop in a canyon to study the map. Standing still under tall buildings is when the position wanders most.
If a route is consistently bad, run it twice with different settings. Comparing two traces from the same place is far more informative than comparing two watches on different days. For the wider picture of how receivers combine constellations, see GPS versus GLONASS versus Galileo.
How to Judge Whether a Trace Is Good Enough
Judge the trace by what you need it for, not by whether it looks pretty.
- Training pace and distance: shape matters less than the total. A 2 percent error on an easy run changes nothing.
- Route sharing and navigation: shape matters. A trace that jumps a block sideways will mislead anyone following it.
- Race analysis: look at splits, not the map. Splits are timed by the watch and are immune to GPS noise.
- Personal records: use a measured course or a track. GPS distance is not a certified measurement and never has been.
The bottom line: expect a clean line in the open, a wobbly one under trees, and a jagged one between towers. If the wobble stays within a few meters and the distance lands within 3 percent, the recording did its job. If it does not, change where you start, how you wear the watch, or the recording mode before you blame the constellation.
Frequently Asked Questions
Why does my watch draw a straight line through a building? That is usually a dropout, not a shortcut. The receiver lost enough satellites to stop producing fixes, so the watch held the last known point or interpolated between the point before the gap and the point after it. The map then draws a straight segment between them. Dense streets and long underpasses cause it most often.
Does dual-frequency GPS fix tree cover? It helps but does not solve it. L5 reception reduces multipath and ionospheric error, so traces under canopy tend to be steadier and recover faster. Foliage still blocks signal, and a watch with a small antenna can still lose lock in dense wet forest. Expect fewer spikes, not a perfect line.
Why is my distance longer in the city than on the same measured route? Noise inflates distance. When each fix lands a few meters off your true path, the watch adds up all those small sideways moves as if you ran them. City routes with reflective buildings produce the most of this. The effect is usually 1 to 5 percent on a short run.
Does waiting before I start really improve accuracy? Yes, especially on a cold start. The receiver needs a clear view to download orbital data and settle on a stable fix. Starting immediately means the first minute of your run is recorded while the watch is still working things out. Thirty to sixty seconds in the open is enough in most conditions.
Is watch GPS less accurate in winter? Indirectly. Bare trees block less signal than leafy ones, which helps. But cold weather can reduce battery voltage and some watches throttle GPS or reduce recording frequency to protect runtime. Snow cover also changes the landscape without changing the map, so traces can look offset even when the position is fine.
See Also
- GPS vs GLONASS vs Galileo: How Wearables Find Your Location
- The Science Behind Outdoor GPS Watches: How They Track Elevation and Location
- How to Use GPS Smartwatches for Accurate Outdoor Tracking
- Stop GPS Drift with the 10 Best Outdoor Running Watches This Year
- How to Solve Inaccurate GPS Tracking with the 10 Best Running Watches in 2026