OpenStreetMap

Accuracy and Precision in GPS Units

Posted by bobwz on 2 November 2020 in English (English).

I’ve spent a fair amount of time looking for the “perfect” consumer device to be used in my on-the-ground surveys that provides better-than-average accuracy without spending thousands on survey-grade equipment. This article series is an attempt to catalogue my experience to those who are interested.

Before diving in, I’d like to review the obstacles to accuracy when surveying using GPS-enabled devices. While this subject may be common knowledge to those familiar, it took me a considerable amount of time to understand some of these points and was a major driving factor in my quest for the “perfect” device.

GPS vs. GNSS

Throughout this article I’ll use the term “GPS” to refer to the systems and satellites that provide longitude and latitude location information to a user. Please note that this is not 100% accurate, however, since “GPS” is a specific satellite (known as a “constellation”) location system among many. In fact, GPS is part of a general Global Navigation Satellite System or GNSS. Other GNSS systems are as follows:

  • GPS
  • GLONASS
  • Galileo
  • BeiDou

I’ll be using “GPS” in an attempt to minimize confusion to those (like myself) who were unfamiliar with this distinction until now.

Precision Limitations

In brief, GPS works by analyzing a time code that is continuously sent by satellites in orbit around Earth. Receivers take that time code from multiple satellites and determine how long it took for the receiver to receive the time code data. This information is then used to calculate a longitude and latitude location on Earth.

Since GPS location is based on calculating the time it takes to beam a signal from space, anything between the GPS antenna and the satellites will impact the time it takes for the signal to arrive and affect accuracy. Buildings, trees, cars, even the atmosphere impacts this timing. This can be seen by logging your GPS position with a smartphone in a field; despite standing still, the position recorded will vary over time giving an idea of precision.

What all this means, is that a GPS signal by itself has a maximum precision. For GPS specifically, the maximum precision is around 5 meters for civilian use. If standing in a field using GPS signals alone to determine your location, you can expect the reported location to be within about 5 meters of your actual physical location on Earth. Add in a small antenna like which is typically found in smartphones and line-of-sight obstructions like buildings, trees, or vehicles, and it is not unreasonable to see the precision degrade to 10 meters or more.

Precision Improvements

To breakthrough this maximum precision limit, receivers must incorporate additional data sources to resolve their location. Most constellations can additionally be augmented by using satellites or ground-based reference stations that feed correction data to receivers. In the case of GPS, these augmentations can increase precision from 5 meters to 2-1.5 meters. Modern smartphones can use cell towers and nearby WiFi hot-spots to improve location data.

Another option is to use two devices. One is positioned at a set location and does not move. The second device (known as a rover) moves about the survey location gathering points of interest. The device at the set location receives GPS information and compares it to it’s set, known location to calculate GPS errors received in the immediate area. With this information, the device can send corrections to the rover gaining precision at the centimeter level. This process, known as real-time kinematic (RTK), is commonly used in professional surveying.

Recently, antenna and receivers have begun popping up that use multiple GPS constellations to refine their precision. Unlike the vast majority of receivers that use one constellation at a time (for example, GPS or GLONASS), these “Multi-band GNSS” receivers will use the data from multiple constellations at a time to determine location. This increases the number of satellites available to calculate the position of the receiver and the number of data points available to help mitigate atmospheric or local interference. I expect these will grow more common over time.

So What?

There are many more topics discussing how GPS does what it does and additional technologies involved, but I wanted to start at this more basic level. It took me a while to understand that GPS accuracy and precision is essentially fixed in the consumer market. I was spending a lot of time shopping around for GPS units without understanding the mechanism in which they function. I thought that buying a more expensive unit or “this years model” I would get better results until gaining centimeter precision. I hope this information reveals that accuracy improvements come from how a device resolves its position, rather than what year it was made, how it looks, or just slapping a “high precision” sticker on the box.

In the next article, I’ll finally get to the devices I mentioned at the beginning and walking through my experience with them.

Comment from philippec on 3 November 2020 at 06:22

The most important thing to consider at this moment is dual frequency GNSS.

Comment from Sanderd17 on 3 November 2020 at 08:13

If you need to go more precise than GPS, your get into land surveyor territory. Land surveyors typically start from a well known point (which is often something precise and high, like church towers), and then perform a mesh of triangulated distance measures to get exact coordinates of some feature (like a parcel of land, or a building).

In Flanders, we actually got access to all surveyed buildings: the front side of the buildings have been measured with cm precision (you can’t go to mm precision, as the walls are most often not that flat).

Comment from bobwz on 3 November 2020 at 13:08

@Samderd17, you’re absolutely correct. For most, gaining access to centimeter precision is extremely expensive and cumbersome for the sake of contributing to OSM. There are some affordable commercially available solutions using RTK and I’ll be talking about those next.

For the most part, my experiences are directed towards those who only are using one device, but still want a decent degree of precision.

Comment from philippec on 3 November 2020 at 17:20

I also use the streetlights and manhole covers for precision. They are or were also in the Flanders official map.

I take perpendicular pictures with house corners, the latter objects, zebra crossings and my POI. I count the pavestones for example.

I know all “buts”, but I go for max.

The “feet” in ID were I handicap, but today I saw it has become meters.

Comment from kucai on 4 November 2020 at 05:09

Even with accurate GPS location, imageries from different providers almost never match each other due to different projection/photo parameters. For example, if we have a known position (wall corner/post box/manhole etc ) that is clearly visible on Bing and Maxar. Then we both realign those imageries to the correct offset. Now, if we go to a point, lets say a building, 1km away, we will find that they do not align anymore despite both were referenced to an accurate, known position.

So, we either needs a grid like reference points or something else. Dual frequency smartphones should be of tremendous help.

My approach has always been to align everything to corrected Bing imagery. Partly because Bing is one of the earliest imagery provider in my area with a lot of OSM object mapped to it. If there’s a newer imagery from Maxar, I’ll find a feature that exist both on Maxar and Bing, and then offset Maxar to Bing.

Personally, I find it mostly to be a minor issue. Mappers using ID wouldn’t have access to the offset db in JOSM, (at least thats how it looks like) and they’ll happily trace with whatever default imagery that ID presents. You could be as precise as you want until somebody moves them again under newer imagery. It becomes a frustrating endless loop. This has been my experience in my corner of the world dealing with paid mappers.


Login to leave a comment