Most search results I get on this talk about the accuracy, that is how close the locations outputted by the GPS are to the real positions, of GPS. However, I'm working on a project that only requires me to know the current 3-dimensional displacement from an arbitrary position, up to an accuracy of a couple centimeters.
Because of this, I just want to know if GPS is precise enough for that job, but it would also be great to know if the error in GPS is usually a constant bias, or mostly random, and if GPS (without RTK or other correction methods) would be sufficient for this kind of job.
Best Answer
Accuracy and precision depend on many factors, including the device and the clarity of the signal received. I'm answering based on the assumption that the device used is not supported by additional signalling equipment on the ground (with known locations).
The signal from the GPS satellites is cleaner if it arrives directly at the device without interference from (degregation due to / reflection from) trees, buildings, mountains, cliffs, etc.
A device may be better or worse at dealing with such interference. A device may be better or worse at calculating a position based on conflicting signals from one or more of the satellites (the device can't necessarily tell which are clean signals and which have suffered as above).
A device may be better or worse at improving the accuracy of the recording of a single point by being left stationary over some time. Some devices can be told that they are stationary (so being able to gather information over time about a single point). Others can't (therefore interpreting changes in signal as movement not additional information)
It's presumably possible, at least theoretically, that a device which is more accurate than others in one scenario might not maintain that increased accuracy (or precision) in other circumstances.
In practical terms, typically on consumer-grade GPS equipment the equipment's calculated location varies, perhaps considerably, each time it calculates. I've come across devices which have been good at precision (noticing a movement of 5 metres despite an overall error of 30 metres - under tree cover) but with no consistency to this. Walking away from the location to an area with clearer signal then returning to the measurement site offsetting the original location by 30 metres.
The more practical extension of your question is probably going to be more like one of the following:
See also: What is the practical limit for lat/long measurement accuracy?