Sect_2.4.2

2.4.2 How Good is GPS?

ERRORS & QUALITY MEASURES

All measurements are subject to errors. Traditionally errors have been classified into three categories:

Random Errors
Systematic Errors
Gross Errors

These are also related to the concepts of precision, accuracy and reliability.

Back To Chapter 2 Contents / Next Topic / Previous Topic

Random Errors

Random measurement errors are, as their name implies, essentially unpredictable (in magnitude and sign) and are basically due to: (a) the "resolution" of the measurement scale (or its "least count"), (b) random internal instrumental effects, and (c) some external but very local effects such as micro-meteorological conditions, electrical connectors and antenna quality, local signal interference, etc. All of these, in essence, define the level of "instrumental noise". In classical statistics the behaviour of random errors can be studied using probability theory, and for most cases in navigation and surveying such errors are assumed to be "white", or have a Gaussian distribution (see HARVEY, 1994, for further reading on this topic).

Systematic Errors

Systematic errors occur according to some pattern, for example:

They may be of constant magnitude and sign.
They may be induced by the instrument, the observer, the physical or environmental conditions.
They may be present in the observation model due to entry of an incorrect parameter or constant.
They may be the result of incorrect application of calibration data.

Gross Errors

Gross errors are the result of blunders or mistakes. If these errors have large magnitudes they are usually easy to identify, and hence can be easily removed. (Another name for gross errors is outliers, and outlier detection is an important step in any navigation or survey data processing procedure.) However, not all gross errors may be large enough to be noticed and hence they can still contaminate the final results.

(Unmodelled or residual biases referred to earlier as belonging to the category of "errors" could be considered as being a type of gross error. Although an unambiguous definition is not possible, in these notes the following convention will be adopted: unmodelled or residual biases which are smaller than the magnitude of the measurement noise are "errors", while those which are larger than the magnitude of the measurement noise will be considered as being "gross errors".)

If systematic and gross errors have been removed, then the accuracy will only be a function of the magnitude of random errors.

Back To Chapter 2 Contents / Next Topic / Previous Topic