Sect_10.4.2

10.4.2 Quality Control Procedures for GPS Networks

SINGLE-SESSION TESTS

Prior to Processing

There are a number of checks that can be made before any data is processed. These can be carried out in the field, or immediately after the observation session's, or day's, data is transferred to the field (or local computation) office. They include the following:

Was the height of antenna measured? Was it checked (by double measurement, or by another person)? Do the values appear reasonable? Was the correct phase centre measured to?
If the antenna was set up on an eccentric station, were the appropriate measurements made to connect the surveyed point to the ground mark?
Does the station log indicate any instrument or power problems? Were the correct satellites (as noted in the log) observed? Were the observation times as planned?
Were all field procedures correctly carried out? Is there any evidence to suggest that the data is in any way unreliable?
Was the correct site occupied? (Check photo evidence, pencil rubbing of ground mark, etc.)
Check system status via electronic bulletin board or information service.

Baseline Processing

The primary source of quality indicators is the GPS data solution itself (including pre-processing for cycle slips, the triple-difference solution, etc.). We can differentiate between preliminary processing for data validation purposes and the final GPS data processing. The strategy and options used for the final processing are often based on the outcomes of the preliminary solutions.

The following are some of the quality information that can be gleaned during the preliminary GPS data reduction phase:

Check the integrity of the data after download, and during reformatting (there may be internal checks for this). Backup copies of the data should be made and verified.
Scan data on a station-by-station basis, to determine the distribution of data to satellites (including the number of obstructions, breaks in data, cycle slips, etc.), the length of the observation session, the satellites tracked (and their health status, etc.). If the common tracking for particular baselines does not satisfy certain criteria (length of session, number of satellites, etc.), a search should be made for station pairs that are best in this regard. The outcome is a station set suitable for processing, as well as initial indications of the best receiver pairings for baseline definition or processing.
Triple-difference solutions, baseline-by-baseline, will provide good apriori coordinates for the final double-difference solution. They can be performed before or after data pre-processing for cycle slip detection and repair. If the data has been "cleaned" prior to the triple-difference solution, any data outliers (data with large residuals) may be uncorrected cycle slips still in the data. The data pre-processing procedures should then be rerun.
If all indications are that the data collection process was trouble-free and the data quality has been validated, the double-difference solution can be attempted.

The above remarks are applied to processing of data collected for conventional static GPS surveys. Because of the short observation sessions and entirely different processing steps followed in the case of high productivity survey techniques such as "rapid static", "stop & go", etc., quality control procedures for data processing are minimal. Instead, the network processing step provides the basis for quality control.

With regards to conventional baseline processing, a number of steps can be taken to satisfy the analyst that the data is indeed sound and the results reliable. Some information useful for quality checks:

Type of solution --> triple-, ambiguity-free or ambiguity-fixed.
The estimate of height difference is of the order of 2-3 times less accurate than the horizontal components.
Input and output coordinates (solve-for and fixed), in various systems, for example Cartesian, geodetic, topocentric. Mean position from single-point pseudo-range solutions.
Echo of receiver (serial numbers, etc.) and station information (site name, antenna height, etc.).
Estimated standard deviation of coordinate components (from variance-covariance matrix).
The correlation matrix or VCV matrix of the coordinate parameters.
Optional estimate of quality of satellite geometry, for example indicators such as "RDOP", etc.
Tracking data acquired, logging times at individual sites, tracking channels used, satellites tracked, signal quality flags, etc.
Number of observations used in solution, as well as those rejected, sampling rate used, data edit criteria, etc.
Summary of ephemeris information, health warning flags in Navigation Message, etc.
Any data reduction performed, for example the tropospheric bias model used.
Indicator of fit of observations to final solution (the residuals) via some indicator such as the "rms tracking".
Possibly the results of statistical tests on the residuals.
If ambiguity-fixed solution attempted, summary of number of ambiguities resolved.
Possibly a summary recommendation as to the quality of the solution.

In general, some of the above information is examined in detail, and an assessment made as to whether they are "reasonable" or "acceptable". If they are "reasonable", that is: they are what is expected from past experience, or they match the manufacturers' specifications, or they pass the requirements set down by the standards and practices for this class of survey, or they satisfy requirements set out in the original contract, the first phase of quality control assessment has been completed.

It must be emphasised that there are no "hard-and-fast" rules. Rarely does GPS reduction software give the analyst access to the observation residuals, and other quantities that would allow extensive statistical tests to be applied. It is therefore important for the analyst to be aware of certain characteristics of double-difference solutions (section 8.1.3) that could be considered "rules-of-thumb". Some of these are:

The chances of successful ambiguity resolution is essentially a function of baseline length, number of satellites tracked, and length of observation session.
The "rms of tracking" increases with increasing baseline length.
The coordinate standard deviations are lower for an ambiguity-fixed solution than for an ambiguity-free solution.
If there are doubts concerning the quality of the ambiguity-fixed solution, it is preferable to accept the ambiguity-free solution in its place.
If the ambiguity-free solution indicates high "rms tracking" values, for example because the baselines are > 50km, the triple-difference solution may be preferable.
Improvement in the modelling of biases (for example, through the use of dual-frequency instrumentation), or sophistication of solution (rigorous adjustment, additional parameters, etc.) leads to better and more reliable results for the same length baseline.

A quality assessment can be made on the basis of the session solution, either derived from single baseline or multi-baseline processing. Some recomputation of the session solution may confirm any suspicions concerning data quality, and even permit the source to be pinpointed. No checks can be made on the consistency of the results, as that requires examination of the multi-session solution. However, where a problem related to data integrity has been identified (or is suspected) in a session double-difference solution, additional "trouble-shooting" can be carried out:

For example, all baselines can be processed separately. It may indicate which of the independent baselines are "best" (those for which the ambiguities were resolved, or the greatest period of common tracking of satellites occurs). These baselines could then be explicitly identified for the final adjustment.
Baselines may be processed in the two directions (fixing first one end of the baseline and solving for the other, and then reversing the process). This could, for example, check the entry of such parameters as the antenna height, especially if the processing were carried out by different analysts.
The processing of the baselines can be carried out twice, by different analysts, preferably using different processing options, to obtain an appreciation of the sensitivity of the GPS solutions to operator subjectivity. Although most processing software has default options to cater for the majority of GPS jobs, check processing using different analysts and options can partially guard against institutional "mind-set" taking hold with regards to GPS data processing.

The outcome is either a satisfactory session solution or an unsatisfactory session solution which cannot be "salvaged" by any amount of re-processing, and which would have to be scrapped. If the quality testing were being carried out as the GPS survey was progressing, it is generally a relatively simple matter to reoccupy the stations and repeat the observations.

Back to Chapter 10 Contents / Next Topic / Previous Topic