Page 150 - Green - Maritime Archaeology: A Technical Handbook. 2nd ed

Chapter 4: Conventional Survey 129
The smaller the residual RMS the better the network fits together. The average absolute residual can also be used but the RMS is preferred as it is more sensitive to single large residuals remaining in the network. Unfor- tunately, the residuals on their own are not a good measure of the fit of the network as they do not include any information about the shape of the network. A poorly designed network can have very small residuals for its measurements but still give large computed position errors for the points. The geometric error is a value showing the overall horizontal position error for the network or how well all of the points have been positioned. Again, the smaller the number the more accurate the survey. The depth error is an equivalent value for the depth axis of the point positions. These figures need to be used alongside the RMS of the residuals when stating the quality and precision of the network. If both the geometric error and the RMS residu- als are smaller than the required specification, then this part of the survey process is completed. In most cases there will be blunders or omissions in the measurements and some additional work is required.
13. Identifying Suspect Measurements
If your RMS residuals and unit variance is large it will be necessary to find the measurements causing the problems and fix whatever is wrong. The figures are a summary of all of the residuals for all of the measurements in the project, so if the RMS is too big then one or more of the residuals is too big.
In the chart window acceptable measurements are shown in green, mea- surements that are too long are shown in blue, and measurements that are too short are shown in red. The thicker the measurement the larger its resid- ual; the thickest have the largest error.
On the observation list window the measurements are shown along with their residuals and their w-test values. Measurements that are too long have positive residuals, those that are too short have negative residuals. In the observation list window you can sort the measurements by column by click- ing on the appropriate column label.
14. Rejecting Measurements
The next step is to identify suspect measurements or blunders with large residuals. The w-test value is a kind of residual and can also be used to identify blunders. Select the most obvious blunder, usually the one with the largest w-test value, and then set it to be ignored. Readjust the network and see if the answer improves. All residuals will be small and the chart will