# How accurate is Google Earth

I always use Google Earth to plan my cross-county routes. Lately, I have begun to ride long distance trips (over 70 kms). Checking my bike computer against Google Earth I have noticed consistent discrepencies of over 10% (computer give me more distance than Google Earth).

I have checked my computer againts my hand held GPS (old Garmin eTrex) and both differ by less than 0.5% in short test (2.5 kms). Also, I cross checked with my ride partner and we both are getting the same readings (+/- 3%) from our computers (different brands) over long distances (20-30 kms).

My bike computer is generic cheap chinese model. I can only select whole wheels size (26"), with no fine tunning. I trace my routes over dirt/paved road, so I certainly keep my planned course.

Can you explain why real distances (according to my computer) are always +10% over Google Earth? What's your experience?

• First and foremost, I would suspect that your cheap Chinese bike computer is the source of the error as Benzo suggested in his answer. – Carey Gregory Jan 16 '15 at 0:23

It largerly depends on how you are marking the route in Google Earth and how you actually ride it. There are a couple of factors that introduce small diferencies when measuring distances and their effects are multiplied over distance, so the longer the route, the bigger the difference.

The straight line effect

For example, if I plan my route drawing a line over the map, but I only mark some points, Google will draw a straight line between two consecutive points. Remember that the shortest path between two points is a stright line, any other path will be longer. But when you actually ride the route you hardly ride straight lines, you ride curves. Each curve will add a little more than each line segment in your map, so, when adding all these diferencies up, the result is a longer actual ride than the planned line.

When following or marking a route using some sort of GPS based solution, you have to account also for the "sampling frequency" of the device. For example, lets say you take a minute to travel around a roundabout. If your GPS device also "samples" your position every minute, it will likely mark your position twice in the same spot, so it will record 0 traled distance. This will affect you more in the more curvy trails and routes, specially if you ride at higher speeds. GPS based solutions also work calculating the distance from two consecutive points and adding up with the next pair, and so on. (Yes, it asumes a straight line between points). The higher the sampling fequency the more acurate the distance will be recorded, but also battery consumption will be higher and if files are a concern, they will also be bigger.

This also occurs in the vertical dimension. For example if you side over a small hill first going up a few meters, then going down over an otherwise straight path, but you marked only the beginning and ending point, most likely the software wont take the hill into account, measuring the distance as the straight line between the points.

The curved line effect

There are also another factor that comes into play: We all know that most bike odometers work by counting wheel revolutions. But many of us do not mentally acount for many wheel turns that are not actually accounting for travel distance. Lets say, you lift up the bike and let the wheel spin, or when you get to a resting point and do a few empty loops while waiting for other riders, etc...

Something else is that, in curves, the front wheel actually travels a longer distance than the rear, and you, actually travel a distance that is in between. For example, lets say you ride trhoug a puddle that makes both your tires very wet, an then on a dry surface you ride a really tight U-turn. Now see the path marked by your tires. The fron one must have drawn a bigger semicircle than the rear one. If you want to be sure, measure them with flexible metric tape ;) This seems too little, but over a long route, every tight corner you ride, actually adds up to the distance "seen" from the front tire.

The longer the route, the more this will affect your predicted route distance versus actual ridden distance. But also, the more curvy routes are the ones more severely affected.

To improve the distance acuracy of the planned route, use shorter line segments along curved parts of the trail, i.e. put the points close together.

Note: I have always doubted about wether Google really takes into account ground level when measuring distance over traced paths, I don't know if it includes altitude of the ground at the marked point or if it assumes all points are at the same height. If it asumes same height, calculated line length is less than the actual path.

• Google Earth measure distance on the horizontal plane (I tested it). However, difference is minimum. Example: a 300 mts high hill with a 4,000 mts long straight road (my tipical ride). Google Earth gives 4.000 mts, but trigonometry say 4,011 mts. Less than 0.3% difference. Also, I draw my path inside existent roads, adding as many points as necesary to stay inside the road. Now, supose we ride tracing a sine wave with 1 meter amplitud (to each side) and 40 mts "wave lenght". So, for a straight 40 mts path, we travel 40.2 mts total. That's a mere 0.5% difference. – user5369 Jan 11 '13 at 19:32
• I also made a practical test. I draw a straigh line between two cities (~16 kms apart). With just 2 points, Google gives 13,240 mts. Then I divide each segment in two equals parts and move new points over the road (rinse & repeat). With 9 points, 13,982 mts. With 33 points, 15,044 mts. Finally, with 65 points, 15,502 mts. Now, when I trace a path keeping segments always inside the road (75 points), "real" distance is 16,300 mts. That is just 5% more than the 65 point all-segments-equal-length path. So, it really doesn't make much difference how precise is the path. – user5369 Jan 11 '13 at 19:48

Since you can't fine tune your wheel size, the bike computer likely is estimating your wheel to be larger than it actually is, causing this discrepency. This is because a larger wheel covers more ground per revolution, so over time that adds up to a sizable difference and probably scales pretty evenly with whatever distance you cover. You may need to get a computer that allows you to measure your wheel distance traveled. See details on this related Question to see how to setup the computer an accurate wheel size.

I think it's hard to estimate your GPS accuracy with a short test. There is always some possible margin of error in using a gps unit, depending on how often it records waypoints and the level of accuracy on each measurement. I'd retry your GPS test over a longer period of time, it may vary more depending on the length of the ride. It could cut off a few meters here and a few meters there, which could eventually add up over the course of a ride.

Google isn't prefect either (as much as I want them to be). It's good to be aware of the margin of error in your measuring systems so you can make accomodations and adjustments as needed.

• An old etrex is pretty good with a clear view of the sky, but even with battery saving disabled will miss a lot of deviations from a straight line. looking at bikecalc.com/wheel_size_math for 26" mtb there about a 12% difference in circumference possible. I would suggest that the OP has rather skinny tyres compared to the bike computer's assumption, causing a loss of distance which coincidentally comes close to the loss due to the GPS skipping some wiggles and rounding off corners. There's often some damping in a GPS trace. – Chris H Jan 16 '15 at 10:26

I wouldn't be surprised to find a 5% discrepancy, though 10% sounds a hair large. Google has actual mileage measurements for major highways, but for minor roads they must estimate. The Google estimates will generally be a bit short, because they don't quite have as many twists and turns and, especially, ups and downs as "real life".

There's also the point that, if you're lazily riding along a country road, you may not keep a straight path but wander back and forth in the roadway, but I assume for you're rides you'd been holding the bike pretty straight.

• Wandering across the lane shouldn't make any significant difference. Even if you're so drunk that you drift 1m to the side for every 10m along the road, you've still only actually covered 10.05m, which is only half a percent. Also, Google are surely factoring in elevation gain; though, again, at the sort of gradients you see on roads, the difference between the "up the slope" distance and the projection onto the plane is tiny. (Actually, the real problem with slopes is the granularity of the height measurements, which is very poor and causes significant inaccuracy.) – David Richerby Oct 8 '17 at 12:38

Google Earth is pretty accurate for straight line distances, for the UK at any rate. I did a small test of this by plotting some grid points from an Ordnance Survey map on Google Earth, and using two-point paths to measure the distances between them. Each section should have been 10 km long. The sum of 24 such sections was 239.999 km, i.e. only one metre out, or 0.0004%, which given the difficulty of precisely locating some of the grid points is not bloody bad!

GPS/Google Maps is more of a map projection which means all the distances are grid distances on a flat plane and your odometer is measuring ground distances on a rolling surface. The ground to grid difference can be corrected by applying a combined scale factor, but the slope distances would be almost impossible to calculate.

• The slope distances are in no way "impossible to calculate". Software which has access to the contour information can calculate the slope distance just as easily as the flat-plane distance. The issue is that contour information is often at rather poor granularity, which makes these calculations inaccurate. – David Richerby Oct 8 '17 at 0:23
• You can get pretty accurate results by just ignoring the contours. You need a 14% climb to get one percent of additional distance compared to horizontally measured and that is already really, really steep. For 10% difference you'd need a 48% climb, steeper than most stairs. – ojs Oct 8 '17 at 17:43