Leica Infinity

Advanced Adjustment
Concepts

Version 2.0 - EN
Table of Contents
1. Before the network adjustment ...................................................................................... 3
2. Explaining the adjustment parameters ............................................................................ 3
3. Adjustment types .......................................................................................................... 10
4. Explaining the network adjustment report .................................................................... 11
5. Frequently asked questions .......................................................................................... 15
6. Proposed workflow for running network adjustments ................................................. 16

Introduction

The purpose of this document is to describe the advanced network adjustment concepts in
Infinity, with more focus on the parameters that affect the network’s reliability.

The first chapter discusses the measurements that are used as observations in network
adjustment.
The second chapter explains the adjustment parameters and the third chapter describes the
supported adjustment types.
In the fourth chapter, a detailed description of the network adjustment report is presented
along with an in-depth explanation of the statistical tests.
The fourth chapter provides answers to frequent questions related to network adjustment
results. The last chapter describes the proposed workflow for running network adjustments.

It is assumed, that the reader is already familiar with the Infinity network adjustments
module. If not, then the “How to…” tutorials related to network adjustments are
recommended.

Leica Infinity, Advanced Adjustment Concepts 2

1. Before the network adjustment
After the data have been imported in the project, the next step is to proceed to the necessary processing
to derive the measurements (referred to as “observations” from now on) that will be used in the network
adjustment.

In case GNSS raw data have been collected, a set of suitable baselines need to be processed first, as these
are the observations that will be included in the network (namely the Δx, Δy and Δz baseline components).
The aim is to organize the field measurements in such a way so that a network of independent baselines
can be formed.

In case TPS raw data have been measured, the observations can either be the raw data themselves or the
result of an application. For example, if data have been collected through Measured Foresights or Sets of
Angles applications, then you can choose whether to use the result of those applications (that is, the
reduced observations) or the raw measurements themselves. This option can be found in the Adjustments
ribbon tab, under Advanced Terrestrial.

In case Level data have been measured, the level lines are the observations that will be used in the network
adjustment. If, however, the level data have not been stored in the instrument, Infinity supports the option
of typing in the height differences directly. This option can be found in the Processing ribbon tab, under
Height Observation.

2. Explaining the adjustment parameters

In this section, the adjustment parameters are explained.

General:

Leica Infinity, Advanced Adjustment Concepts 3

In Computation, you can select whether the adjustment results will be stored immediately after
computing. This option is unchecked by default, as the normal case is that there will be many adjustment
runs before reaching the desired result.
The “Control Points” option determines whether the control points will be treated as absolutely
“Constrained” or “Weighted”:

• Select “Constrained”, if you want to keep the control points fixed (ie without receiving any
corrections) during adjustment. Their standard deviations will be set to zero.

• Select “Weighted”, if you want the control points to receive corrections during adjustment. Their
standard deviations will be the ones you set.

Notes:
1. Fixing the control point coordinates often imposes some distortion (scale) on the network.
2. At least three control points need to be available in the project, otherwise treating the control
points as “Weighted” will have the same impact on the result as if they were set as
“Constrained”.

In the Iterations settings, you can set the maximum number of iterations as well as the threshold
correction value for the iterations. It is suggested to keep the default values.

Leica Infinity, Advanced Adjustment Concepts 4

The Visualisation Exaggeration Factors refer to the visualisation of the accuracy of the network in terms
of Error Ellipses (Ellipsoids in 3D View) and Reliability. The bigger the values entered, the bigger the
graphical element will be drawn. If the ellipses or the reliability are smaller than the Threshold, they will
not be drawn in the view.

The Confidence Levels can be modified separately for the error ellipses and the heights. The confidence
levels vary from country to country and can be set up to 99%. The bigger the confidence levels, the bigger
the axes of the error ellipses (or ellipsoids in 3D view).

TPS Accuracy Information:

In Standard Deviations, the standard deviations for the observed angles and distances can be set. You
can also define whether individual or default values for all observations will be used.

In Centring/Height Errors, the centring and height errors for setup and target points can be set. You can
also define whether individual or default values will be used.

Notes:
1. Setting the proper standard deviations as well as realistic centering and height errors is of very
high importance, as these values have an immediate impact on the weight matrix of the
Leica Infinity, Advanced Adjustment Concepts 5
 observations.
2. The more you increase the standard deviations and centering and height errors, the less the
weight of the observations will be and the less sensitive to error detection the network
becomes.

GNSS Accuracy Information:

Similar to TPS Accuracy Information, the GNSS Accuracy Information includes standard deviation and
centring and height errors. If the Source for Standard Deviations is set to “Individual”, each baseline is
weighted according to the variance-covariance matrix that was computed along with the processed
baseline. If “Use Defaults” is selected, then all the baselines will be weighted relative to their length, using
the Absolute GNSS Baseline and Relative GNSS Baseline user-entered values.

In many cases, the variance-covariance matrices of a processed baseline are too small and are considered
as too optimistic. Before running an adjustment, it is recommended that the baselines acquire a
reasonable weight. The Sigma a priori (GNSS) re-scales the variance-covariance matrix for each baseline
by the user-entered value. Over-rescaling of the Sigma a priori (GNSS) will make the network less sensitive
to error detection.

Leica Infinity, Advanced Adjustment Concepts 6

Level Accuracy Information:

As with TPS Accuracy Information and GNSS Accuracy Information, the Level Accuracy Information
includes settings for Absolute and Relative Level Line standard deviations as well as level observations.
Assigning the proper weight to each level line observation is a key element to obtaining realistic
adjustment results.

Test Criteria:

To estimate the reliability of the network, Infinity uses the B-Method of testing. The Level of Significance
is the probability of committing a Type-I error, whereas the Power of Test is the power of the test against
a Type-II error.
Regarding the Sigma a posteriori, there are three cases:
1. It can be used to rescale the final variance-covariance matrices and must be handled with care,
especially if its value is significantly greater or smaller than 1.0.
2. It can be used only if the F-Test fails and in this case, there is always a general problem in the
adjustment that needs to be investigated
3. It can be ignored. This will keep the computed variance-covariance matrices unscaled and it is a
good choice for the initial inner-constrained adjustment runs.

Leica Infinity, Advanced Adjustment Concepts 7

Advanced Terrestrial:

In case reduced observations are preferred over raw TPS data, the Use reduced observations setting must
be enabled. The standard deviations of the reduced observations will then be used to compute the weight
matrix.
In some cases, it is worth extending the mathematical model of the adjustment, as this can make it more
suitable to fit the observations. Two parameters are commonly used: the Vertical Refraction Coefficient
and the Scale Factor Correction.
The Vertical Refraction Coefficient can be useful, if small TPS 3D networks are measured. This parameter
affects the measured zenith angles.
The Scale Factor Correction is useful when GNSS and TPS observations are used in a combined adjustment.
More details on this will be given in the next chapters.
Each of these parameters can be ignored, computed or used with a value that is user-entered. This
provides flexibility that in many cases is important.

Leica Infinity, Advanced Adjustment Concepts 8

Coordinate System:

One of the most important aspects of network adjustments is to select the coordinate system which the
observations will be referred to. This is something different from the coordinate system that is used in
the project. The latter is clearly used to project the point coordinates to it, whereas the first is used to
transform the observations from different sources, so that they can be properly combined. As an example,
we take the case where GNSS and TPS observations need to be used in a combined adjustment. The GNSS
observations are referred to a geocentric coordinate system, whereas the TPS observations are referred
to a topographic plane. In order to use the two observation groups in a combined adjustment, it is
necessary to transform both groups to a common reference. Since the GNSS observations are treated as
true 3D vectors, we choose to transform the TPS observations from the topographic plane to the same
geocentric system the GNSS observations are referred to. This geocentric system is the “WGS84” option
in the Coordinate System dialog box.

The “Local Grid” option is a good choice when only TPS data are adjusted or when TPS and Level data are
combined. In this case, the GNSS data will not be used and a warning message will appear.

The “Local Geodetic” is a good choice when GNSS, TPS and Level data need to be used in a combined
adjustment, provided that there are at least 3 control points whose coordinates are known in this “Local
Geodetic” system. This option offers great flexibility as it extends the mathematical model of the
Leica Infinity, Advanced Adjustment Concepts 9
adjustment so that it includes additional rotation and scale parameters to transform the observations to
this “Local Geodetic” system.
Notes:
1. A local coordinate system needs to be set as Master in the project in this case.
2. The transformation used in this local coordinate system must be either a Classical 3D or None
for this option to become active.

The additional rotation and scale parameters can either be computed or used with a user-entered value.
In addition, Ellipsoidal or Orthometric heights can be used. This is very important when GNSS and Level
data are combined. The “WGS84” or the “Local Geodetic” (in case enough control points and a suitable
coordinate system are available) option could be used as the coordinate system for the adjustment.
In this case, the orthometric height differences will be transformed via the geoid to ellipsoidal height
differences; they will be adjusted in WGS84 (or the Local Geodetic) and then transformed back to adjusted
orthometric height differences to compute correct orthometric heights for the points.

3. Adjustment types
Depending on the dimension of the data, Infinity supports the following types of adjustment:
1. 3D (Full) adjustment: this type can be used to adjust pure GNSS data or a small TPS 3D network or
combined GNSS, TPS and Level data.
2. 2D adjustment: this type is more suitable to adjust TPS data ignoring the height dimension.
3. 1D adjustment: this type can be used to adjust pure level data or combined level and trigonometric
height differences.

Infinity can also run adjustments in two steps. Two additional adjustment types are supported: 2D+1D and
1D+2D. When TPS and level data are combined, the point positions can be computed from the TPS
observations (2D position), whereas the point heights can be computed from the level data (1D position).
Thus, the trigonometric height differences can be ignored, if necessary.

Depending on the constraints imposed on the data, Infinity supports three types of adjustment: “Inner
Constrained”, “Minimally Constrained” and “Constrained”.

The “Inner Constrained” option uses a special type of minimum constraints called inner constraints. In this
case, the corrections the point coordinates receive are the smallest possible. Also, all control points are

Leica Infinity, Advanced Adjustment Concepts 10

considered as unknowns.

The “Minimally constrained” option uses user-defined minimum constraints. In this case, the corrections
the point coordinates receive are smaller than the ones computed in a constrained adjustment. An
example of minimally constraining a network is to fix one control point in height and adjust level data in
1D, or to fix one control point in position and height and adjust GNSS data.

“Constrained” is the case that all or more than the minimum required control points are kept fixed in the
adjustment.

When more than the necessary constraints are imposed on the data, some distortions are expected in the
network. This causes some bias in the coordinate estimations and it is for this reason that this type of
adjustment should be avoided, when checking for possible outliers in the observations.

4. Explaining the network adjustment report

After running an adjustment, a report can be created from the Reports ribbon bar button.
The network adjustment report comprises the following sections:

• Project Details

• Adjustment Settings

• Adjustment Summary

• Input Data

• Adjustment Results
• Testing and Estimated Errors

Project Details:
All the project details can be found in this section: General, Customer details, Master coordinate system,
etc.

Adjustment Settings:
This section includes all the necessary information on settings used in adjustment. In summary,
information on the control point treatment, the adjustment dimension and type, the coordinate system
the observations are referred to, the height mode as well as the standard deviations and the testing
criteria is displayed. There is also information on whether reduced observations are used or not.

Leica Infinity, Advanced Adjustment Concepts 11

Adjustment Summary:
This section includes all the necessary information regarding the adjustment summary. Total number of
unknowns, of observations and of constraints as well as degrees of freedom, optimization criterion and
sigma a posteriori can be found here. The F-Test and Chi Square Test actual and critical values are included
here along with the critical values for W- and T-Tests.

Notes:
1. If the calculated value of the F- (or the Chi Square) Test exceeds its critical value (upper or
lower boundary respectively), then it will be marked in bold red.
2. The sigma a posteriori is the square root of the result of dividing the optimization criterion
(v’Pv) by the degrees of freedom.

Input Data:
This section includes the approximate values for the point coordinates as well as the initial observations
with the standard deviations. The standard deviations are used in the weight matrix.

Adjustment Results:
This section includes the adjusted point coordinates with their standard deviations, the absolute error
ellipses, the external reliability values and the adjusted observations. If GNSS baselines have been included
in the adjustment, the baseline vector residuals can also be found here. Next to each observation, the W-
Test values (and T-Test values for 3D observations) are shown.

Note:
By default, the error ellipses are calculated at 39.4% confidence level, whereas the 1D-
coordinate elements are calculated at 68.3%.

Testing and Estimated Errors:

This section includes important information on the reliability of the network. For all kinds of observations,
the table of Observation Tests is available. Reliability is divided in two kinds:

• Internal reliability, which is expressed by the Minimal Detectable Bias (MDB). The MDB presents
the size of the smallest possible observation error, still detectable by the statistical test (data

Leica Infinity, Advanced Adjustment Concepts 12

 snooping) with a probability equal to the power 1-Beta of the test. A large MDB indicates a weakly
checked observation or coordinate. Thus, the larger the MDB the poorer the reliability. If an
observation is not checked at all, no MDB can be computed and the observation is marked as a
'free observation'.

• External reliability is expressed by the Bias to Noise Ratio (BNR). The external reliability is used
as a measure to determine the influence of a possible (undetected) error in the observations on
the adjusted coordinates. The BNR of an observation reflects this influence, whereby the size of
the observation error is defined equal to the MDB of that particular observation. The BNR is a
dimensionless parameter combining the influence of a single observation on all coordinates. The
BNR can be interpreted as the ratio between reliability and precision.
An important property of both the MDB and BNR is that they are independent of the choice of the control
points.

The Red expresses the redundancy (%) of each observation. The higher the redundancy the more
controllable this observation is. Theoretically, the redundancy of an observation between two control
points that are kept fixed is 100, whereas the redundancy of a free observation is 0.

The W-Test is a one-dimensional statistical test that is suitable for detecting possible outliers. It’s critical
value depends on the Level of Significance.

Essential for the B-method of testing is that an outlier is detected with the same probability by both the
F-Test and the W-Test. For this purpose, the power 1-Beta of both tests is fixed on a level of usually 0.80.
The Level of Significance Alpha0 of the W-test is also fixed, which leaves the level of significance Alpha
of the F-Test to be determined. Having Alpha0 and 1-Beta fixed, Alpha depends strongly on the
redundancy in the network. For large scale networks with many observations and a considerable amount
of redundancy, it is difficult for the F-Test to react on a single outlier. The F-Test, being an overall model
test, is not sensitive enough for this task. Because of the link between the F-Test and the W-Test by which
the power is forced at 0.80, the level of significance Alpha of the F-test will increase. Considering the
above, it is common practice to always carry out the data snooping, no matter the outcome of the F-test.

The W-Test works well for single observations, e.g. directions, distances, zenith angles, azimuths and
height differences. However, for some observations such as GNSS baselines, it is not enough to test the
DX-, DY-, DZ-elements of the vector separately. It is imperative to test the baseline as a whole as well.
Therefore, the T-Test is introduced. Depending on the dimension of the quantity to be tested, the T-Test

Leica Infinity, Advanced Adjustment Concepts 13

is a 3- or 2-dimensional test. As with the W-Test, the T-Test is also linked to the F-Test by the B-method
of testing. The T-Test has the same power as both F- and W-Tests, but it has its own level of significance
and its own critical value.

The Estimated Error is the size of the error responsible for the rejection of an observation or known
coordinate. It is a useful tool, yet it should be handled with care:

• Only the estimated error associated with the largest W-test or T-test should be considered.

• As far as the W-test is concerned, the Estimated Error is based on the conventional alternative
hypothesis that just one observation or known coordinate contains an error. Consequently, if more
errors are present in the network, the result of the estimation could be meaningless, unless errors
have been made (geographically) far apart.

• As far as the T-test is concerned, the Estimated Error is based on the hypothesis that just one
GNSS baseline or known station contains an error. Consequently, if more errors are present in the
network the result of the estimation could be meaningless, unless errors have been made
(geographically) far apart. The test results and Estimated Errors are only meaningful when
observational errors have been filtered out in the foregoing free adjustment and testing phase.
This is the reason that you always need to run an inner- or a minimally constrained adjustment
before adding the control points.

In case GNSS observations are adjusted, a special test that checks the antenna heights for gross errors is
available in the table of Antenna Height Tests.

Notes:
1. It is desirable that the BNR is homogeneous for the entire network.
2. The MDB and the BNR values are affected by the selection of the weight matrix.
3. Always strive to retain sensibly small MDB values or else the network might become very
insensitive in detecting outliers.
4. It is advised that the Red value exceeds 20 for all observations.
5. There might be cases that an observation is marked as a possible outlier, but the estimated
error is small –even smaller than the MDB value for this observation. In such cases, it is up to
you to decide whether this observation is really an outlier or not.

Leica Infinity, Advanced Adjustment Concepts 14

5. Frequently asked questions
When running an adjustment, the user may come across some values that are out of tolerance or
unacceptably high. Some suggestions are provided as to what the issue may be and how to deal with it.

1. The F-Test failed or is way lower than the critical value

This means that either there is a weighting issue with the observations or that there are possible
outliers among the observations, or both. This case requires further investigation.

• If the F-Test is much lower than its critical value, this means that there is a weighting
issue. Consider changing the standard deviations of the observations and/or the control
points to be more optimistic.

• If the F-Test is close to its critical value, then it failed due to some outliers lying among
the observations. The outliers must be removed. In this case, if the F-Test failed during an
inner-constrained adjustment, consider removing the outlying measurements. However, if
the F-Test passed in an inner- or minimally constrained adjustment, but then failed in the
constrained adjustment, consider checking the control point coordinates for gross errors
or standard deviations that are too optimistic.

• If the F-Test is much higher than its critical value and almost all W-Tests/T-Tests fail, then
the assigned weights are too optimistic. Consider relaxing the standard deviations of the
observations and/or the control points.

2. The Chi-Square test failed

Like the F-Test failing case, if the Chi-Square test is much lower than the lower bound, this
clearly indicates a weighting issue. If the test is higher than the upper bound, this could mean
that either the assigned weights are too optimistic or that there are outliers among the
observations, or both. Follow the same procedure as when the F-Test fails.

3. The External Reliability table shows huge values

This is probably due to external information for position missing. Consider adding control points
and running a constrained adjustment.

4. The W-Test/T-Test for a control point failed

Check the coordinates of the control point for gross errors. Also, check if the standard deviations

Leica Infinity, Advanced Adjustment Concepts 15

 of the coordinates are too optimistic.

5. The W-Test/T-Test for some observations failed

Check the failing observations for gross errors and remove only one observation from the
adjustment: the one that has the highest absolute value for W-Test or the highest value for the
T-Test.

6. I remove one outlier at a time and in the next adjustment run a new one pops up
This may happen if you have set sigma a posteriori to “Use”. This option re-scales the Qxx matrix
of the results and makes the W-Test/T-Test stricter than it should probably be. If after setting
the sigma a posteriori to “Ignore” the issue persists, this means that the network is not
geometrically consistent. Consider adding more observations and checking the scale factor,
especially when GNSS and TPS observations are used in a combined adjustment.

6. Proposed workflow for running network adjustments

The proposed workflow for running network adjustments is shown in the following chart:

Leica Infinity, Advanced Adjustment Concepts 16

 Run Inner Constrained Adjustment

Yes Change Weights

Yes
Almost all (Make sure MDB values remain small
F-Test fails
obs marked enough)

No
No

Yes Yes
Blunders are Estimated Remove 1 observation with max absolute T-
detected Errors are big or W- test value

No No

Add all Control Points

Run Constrained Adjustment

Yes
Remove Control Point with max absolute T-
F-Test fails
or W- test value and with big Estimated
Error

No

Store Result

Leica Infinity, Advanced Adjustment Concepts 17

 Original text
Published in Switzerland
© 2021 Leica Geosystems AG, Heerbrugg, Switzerland

Leica Geosystems AG
Heinrich-Wild-Strasse
CH-9435 Heerbrugg
Switzerland
Phone +41 71 727 31 31
www.leica-geosystems.com