Even though the specifications for a project may allow lower accuracies, the presence of blunders in observations is never acceptable. Thus, an important question for every surveyor is, “How can I tell when blunders are present in the data?” In this chapter, we begin to address that question, and in particular, stress traverse analysis. This topic is discussed further in Chapters 21 and 25.
In Chapter 6, it was shown that the estimated error in a function of observations is dependent on the individual errors in the observations. Generally, observations in horizontal surveys (e.g., traverses) are independent. That is, the measurement of a distance observation is independent of the azimuth observation. But the latitude and departure of a line, which are computed from the distance and azimuth observations, are not independent. Figure 8.1 shows the effects of the errors in distance and azimuth observations on the computed latitude and departure. In Figure 8.1, it can be seen that there is correlation between the latitude and departure; that is, if either distance or azimuth observation changes, it causes changes in both latitude and departure.
Because the observations from which latitudes and departures are computed are assumed to be independent with no correlation, the SLOPOV approach [Equation (6.16)] can be used to determine the estimated error in these computed values. However, for proper computation of estimated errors in functions that use these computed values (i.e., latitudes and departures), the effects of correlation must be considered, and thus the GLOPOV approach [Equation (6.13)] will be used.
When computing the latitude and departure of a line, the following well-known equations are used:
where Lat is the latitude, Dep the departure, Az the azimuth, and D the horizontal length of the line. To derive the estimated error in the line's latitude or departure, the following partial derivatives from Equation (8.1) are required in using Equation (6.13).
Equation (8.1) is based on the azimuth of a course. However, in practice traverse azimuths are normally computed from observed angles rather than measured directly. Thus, another level of error propagation exists in calculating the azimuths from angular values. In the following analysis, consider that angles-to-the-right are observed and that azimuths are computed in a counterclockwise direction successively around the traverse using the formula
where AzC is the azimuth for the current course, AzP the previous course azimuth, and θi the appropriate interior angle to use in computing the current course azimuth. By applying Equation (6.18), the error in the current azimuth, AzC, is
In Equation (8.6) is the estimated error in the appropriate interior angle used in computation of the current azimuth, is the estimated error in the azimuth of the previous course, and is the estimated error in the course being computed. This equation is also valid for azimuth computations going clockwise around the traverse. The proof of this is left as an exercise.
From elementary surveying, it is known that the following geometric constraints exist for any closed polygon-type traverse:
Deviations from these conditions, normally called misclosures, can be calculated from the observations in any traverse. Statistical analyses can then be performed to determine the acceptability of the misclosures and check for the presence of blunders in the observations. If blunders appear to be present, the measurements must be rejected and the observations repeated.
However, while the initial azimuth will correctly orient the traverse, its error does not affect the linear misclosure of the traverse. This can also be stated about the observed angle at the control station, since the azimuth for the first course is either given or assumed. Thus, the errors in these observations can be ignored when checking the linear misclosure of the traverse. The error in the initial angle should be used in checking the angular misclosure in the traverse. The following example illustrates methods of making these computations for any closed polygon traverse.
In Example 8.2, we failed to reject the null hypothesis; that is, there was no statistical reason to believe there were errors in the data. However, it is important to remember that this does necessarily imply that the observations are free from error. There is always the possibility of a Type II error. For example, if the computations were supposed to be performed on a map projection grid,2 but the observations were not reduced, the traverse would still close within acceptable tolerances. However, the computed results would be incorrect since all the distances would be either too long or too short. Another example of an undetectable systematic error is an incorrectly entered EDM-reflector constant (see Problem 2.23). Again, all the observed distances would be either too long or too short, but the traverse misclosure would still be within acceptable tolerances.
Surveyors must always be aware of systematic instrumental errors and follow proper field and office procedures to remove these errors. As discussed, simply passing a statistical test does not imply directly that the observations are error or mistake free. However, when the test rejects the null hypothesis, only a Type I error can occur at an α level of confidence. Dependent on the value of α, rejection of the null hypothesis is a strong indicator of problems within the data.
As illustrated in Figure 8.3, a link traverse begins at one station and ends on a different one. Normally, they are used to establish the positions of intermediate stations, as in A through D of the figure. The coordinates at the endpoints, Stations 1 and 2 of the figure, are known. Angular and linear misclosures are also computed for these types of traverse, and the resulting values used as the basis for accepting or rejecting the observations. In a link traverse, the error in the initial azimuth and angle are important in checking both the angular and linear misclosure of the traverse. Furthermore, the error in the closing azimuth is considered when checking the angular misclosure. It does not affect the linear misclosure of the traverse. Example 8.3 illustrates the computational methods.
This example leads to an interesting discussion. When using traditional methods of adjusting link traverse data, such as the compass rule, the control is assumed to be perfect. However, since control coordinates are themselves derived from observations, they contain errors that are not accounted for in these computations. This fact is apparent in Equation (8.21) where the coordinate values are assumed to have no error and thus are not represented. These equations can easily be modified to consider the control errors, but this is left as an exercise for the student.
One of the principal advantages of the least squares adjustment method is that it allows application of varying weights to the observations, and control can be included in the adjustment with appropriate weights. A full discussion of this subject is presented in Section 21.7.
The computations in this chapter can be challenging or tedious at best to perform using a calculator. Thus, it is recommended that these computations be programmed in the software. On the companion website for this book, an Excel® spreadsheet (Chapter 8.xls) for Examples 8.2 and 8.3 is available for study. Programming allows for an easier and more reliable method of developing the required matrices. Once the initial matrices are computed, they can be manipulated in software such as MATRIX, which is on the companion website, or in the spreadsheet itself.
A spreadsheet often provides routines to manipulate the matrices. In Excel®, matrices can be multiplied using the MMULT() function. Similarly, matrices can be transposed using the TRANSPOSE() function. These functions are demonstrated in the spreadsheet CHAPTER 8.xls. To simplify the selection of the matrices, the block of cells for each matrix can be “named.” The help file that accompanies Excel® discusses how to name cells or groups of cells and use the matrix functions.
For those who wish to develop a more robust program, a higher-level programming language or Mathcad can be used. ADJUST demonstrates this in its traverse computations option. The dialog box for this option is shown in Figure 8.4, where a polygon traverse with angles is selected. At the bottom of the box is the option to compute the estimated errors in the traverse. Selecting this option informs the software to not adjust the angles in the traverse, but rather, to use the entered distance and angle uncertainties to compute the misclosures in both the angles and the traverse. Figure 8.5 depicts the file used in computing the estimated errors for Example 8.2. Notice that with a polygon traverse there is no error given for the initial azimuth of 0°00′00″, and that the estimated errors in the distances and angles follow the observations for each course of the traverse. The help file that accompanies ADJUST describes the format of this file.
This chapter discussed the propagation of observational errors through traverse computations. Error propagation is a powerful tool for the surveyor, enabling an answer to be obtained for the question, “What is an acceptable traverse misclosure?” This is an example of surveying engineering. Surveyors are constantly designing measurement systems and checking their results against personal or legal standards. The subjects of error propagation and detection of observational blunders are discussed further in later chapters.
Note: Partial answers to problems marked with an asterisk can be found in Appendix H.
Backsight | Occupied | Foresight | Angle | S (″) |
A | B | C | 98°06′59″ | ±2.8 |
B | C | D | 85°56′57″ | ±2.6 |
C | D | A | 92°33′50″ | ±3.0 |
D | A | B | 83°22′19″ | ±3.2 |
Backsight | Occupied | Foresight | Angle | S (″) |
A | B | C | 82°42′10″ | ±5.6 |
B | C | D | 68°43′20″ | ±5.3 |
C | D | A | 93°03′30″ | ±4.4 |
D | A | B | 115°31′15″ | ±4.8 |
From | To | Distance (ft) | S (ft) |
A | B | 247.86 | ±0.012 |
B | C | 302.49 | ±0.012 |
C | D | 254.32 | ±0.012 |
D | A | 319.60 | ±0.012 |
From | To | Distance (ft) | S (ft) |
A | B | 256.69 | ±0.012 |
B | C | 303.45 | ±0.012 |
C | D | 326.99 | ±0.012 |
D | A | 160.17 | ±0.012 |
Distance Observations | Angle Observations | ||||||||
From | To | Distance (m) | S (m) | BS | Occ | FS | Angle | S (″) | |
W | X | 180.395 | ±0.005 | Mk1 | W | X | 157°39′42″ | ±3.6 | |
X | Y | 146.622 | ±0.005 | W | X | Y | 96°27′11″ | ±3.8 | |
Y | Z | 251.526 | ±0.006 | X | Y | Z | 172°52′34″ | ±3.7 | |
Y | Z | Mk2 | 185°53′49″ | ±3.1 |
Control Azimuths | Control Stations | ||||||||||||||||||||||
From | To | Azimuth | S (″) | Station | Northing (m) | Easting (m) | |||||||||||||||||
Mk1 | W | 345°57′50″ | ±9″ | W | 5,000.000 | 10,000.000 | |||||||||||||||||
Z | Mk2 | 58°51′09″ | ±8″ | Z | 5079.412 | 10,434.819 |
Distance Observations | Angle Observations | ||||||||
From | To | Distance (ft) | S (ft) | BS | Occ | FS | Angle | S (″) | |
W | X | 223.59 | ±0.012 | Mk1 | W | X | 171°52′06″ | ±5.5 | |
X | Y | 854.40 | ±0.013 | W | X | Y | 137°07′03″ | ±5.9 | |
Y | Z | 460.95 | ±0.012 | X | Y | Z | 290°02′35″ | ±3.4 | |
Y | Z | Mk2 | 77°44′55″ | ±4.8 |
Control Azimuths | Control Stations | ||||||||||||||||||||||
From | To | Azimuth | S (″) | Station | Northing (ft) | Easting (ft) | |||||||||||||||||
Mk1 | W | 251°33′54″ | ±5.0″ | W | 1000.00 | 1000.00 | |||||||||||||||||
Z | Mk2 | 28°21′00″ | ±3.0″ | Z | 1850.00 | 1600.00 |
The EDM instrument has a specified accuracy of ± (3 mm + 3 ppm).
Distance Observations | Angle Observations | ||||||||||
From | To | Distance (ft) | BS | Occ | FS | Angle | |||||
1 | 2 | 999.99 | 5 | 1 | 2 | 191°40′12″ | |||||
2 | 3 | 801.55 | 1 | 2 | 3 | 56°42′22″ | |||||
3 | 4 | 1680.03 | 2 | 3 | 4 | 122°57′10″ | |||||
4 | 5 | 1264.92 | 3 | 4 | 5 | 125°02′11″ | |||||
5 | 1 | 1878.82 | 4 | 5 | 1 | 43°38′10″ |
Control Azimuths | Control Stations | |||||
From | To | Azimuth | σ (″) | Station | Easting (ft) | Northing (ft) |
1 | 2 | 216°52′11″ | ±3 | 1 | 1000.00 | 1000.00 |
Compute: