Chapter 2

General Principles of Surveying

Surveying has two notable characteristics: the work is done to a much

higher level of accuracy than most other engineering work; and it is possi-
ble for quite serious errors to remain undetected until it is too late to correct
them. For this reason, there are some inherent principles which should be
observed in all surveying, regardless of the type of survey or the equipment
used. This chapter describes those principles.

2.1 ERRORS

All the results of surveying are based on measurements, and all measure-
ments are subject to errors. Because surveying involves high degrees of
accuracy (most surveying measurements are accurate to within 10 parts
per million and some are within 1 or 2 parts per million), it is relatively
easy to make significant errors, and relatively hard to detect them. The
understanding and management of errors is therefore possibly the single
most important skill that a professional surveyor must possess. Many of
the techniques of surveying are directed towards cancelling or eliminating
errors, and towards ensuring that no serious error remains undetected in
the final result. Even so, the presence of unnoticed ‘systematic’ errors in a
survey can lead to false, yet seemingly consistent, results. A recent inter-
national tunnelling project drifted several metres from its intended path
because temperature gradients near the tunnel wall caused lines of sight to
bend consistently in one direction, and this was not detected until an inde-
pendent method was used to check the work.
High accuracy in surveying is expensive because it involves costly, high-
quality equipment and more elaborate procedures for taking measurements.
On the other hand, cheaper equipment may not be adequate to achieve the
required accuracy, particularly if (for instance) a long distance has to be
split into several steps, requiring more measurements and resulting in an
accumulation of errors. Surveys are therefore often conducted using high-
quality equipment to establish a few ‘major control’ stations around the

7
8 Plane and Geodetic Surveying

area to a higher accuracy than is required overall, and then filling in the
intervening detail by cheaper methods ade­quate for the shorter distances.
This is usually the most economic way of distributing the ‘error budget’ to
achieve a satisfactory final result at minimum cost.

2.1.1 Types of Errors

Surveying errors fall into three categories:

1. Gross errors. Gross errors are due to mistakes or carelessness, such as

misreading by a metre or a degree. A proper routine of checks should
detect them. A surprisingly common source of error is the manual
transcription of readings from one place to another.
2. Systematic errors. Systematic errors are cumulative and due to some
persistent cause—generally in an instrument, but sometimes in a habit
of the observer. They can be reduced by better technique but not by
averaging many readings, as they are not governed by the laws of prob-
ability. Thus all distances measured with an inaccurate tape or electro-
magnetic distance measurement (EDM) device will, from that cause,
have the same percentage or absolute error, whatever their lengths and
however many times they are measured; the only remedy is to calibrate
the device more carefully. This is the most serious sort of error, and the
technique of survey is mainly directed against it: the greater the accu-
racy required, the more elaborate and expensive the instruments and
the technique.
  A special type of systematic error is a periodic error, which var-
ies cyclically within the instrument. Examples include errors in the
positions of the angle markers on a horizontal circle, inaccuracies
in a vernier scale, or nonlinearities in the phase resolver of an EDM
or global navigational satellite system (GNSS) receiver. This type of
error can sometimes be eliminated by special observation techniques:
for instance, measuring a horizontal angle several times, but using a
different part of the horizontal circle on each occasion.
3. Random errors. Random errors are due to a number of small causes
beyond the control of the observer. Their magnitude depends on the
quality of the instrument used and on the skill of the observer, but
they cannot be corrected. Thus no one can place a mark, or make an
intersection, or read a scale with absolute accuracy or consistency.
Even after allowing for systematic personal bias (covered in ‘Systematic
errors’ above), there will remain errors which are a matter of chance
and are subject to the laws of probability. In general, positive and nega-
tive errors are equally probable; small errors are more frequent than
large ones, and very large random errors do not occur at all.
 General Principles of Surveying 9

  In statistical terms, random errors cause readings to deviate from

the correct value in the manner of a normal distribution—similar, for
instance, to the scatter of heights to be found in a sample of adults.
The scale of the scattering can therefore be defined by quoting the
standard deviation (σ) of the distribution; two-thirds of all readings
will lie within one standard deviation of the correct value (above or
below), and 95% within two standard deviations. Alternatively, the
standard deviation for the random element in an observation can be
estimated by taking the measurement several times, and seeing what
range of values covers the middle two-thirds of the readings; the size
of this range is an estimate of (2 × σ).
  Two other measures of quality are also used to define the accuracy
of readings affected by random errors. The probable error, expressed
as ±p, is such that 50% of a large number of readings differ from the
correct value by less than p; for normally distributed errors, p is 0.675
times the standard deviation of the readings. A more useful measure
of accuracy is the 95% confidence value which, as explained above,
is almost exactly two standard deviations.
  Assuming that the observation errors from an instrument have a
normal distribution (i.e., that they contain no gross or systematic
errors), it can be shown that the standard deviation associated with
the arithmetic mean of a set of n repeated observations is 1 n times
the standard deviation of a single observation. Thus if a single angle
measurement can be read to one second of arc, the mean of four read-
ings should have a precision of 0.5 seconds. Taking the same mea-
surement several times can therefore be a valid way of increasing the
overall accuracy of a survey.

2.1.2 Precision and Accuracy

In understanding the nature of measurement errors, it is important to
appreciate the distinction between precision and accuracy. It is possible,
for instance, to measure a distance to fairly high precision (0.5 mm or bet-
ter) using a simple tape measure—but if the marks have not been printed
in the right places on the tape, the reading will not be accurate. Even when
the greatest precautions are taken in making a reading (e.g., measuring the
distance again, using a different part of the tape), systematic errors (e.g., the
whole tape has become longer because of thermal expansion) may still
domi­nate the results. Caution must therefore be used when estimating the
standard deviation of a set of observations from the apparent ‘scatter’ of the
results, as described above. A set of consistent readings indicates a consis-
tent instrument and a good observer, but not necessarily an accurate result.
It is good practice to avoid recording observations to a higher precision
than is warranted by their accuracy. However, this is not always done, and
10 Plane and Geodetic Surveying

surveyors should be aware that observational data is sometimes consider-

ably less accurate than it appears to be.

2.2 REDUNDANCY

Given two points whose positions are known, the position of a third point
in plain view can be found by (for instance) measuring the horizontal dis-
tances between it and the two known points. However, the accuracy of the
calculated position can only be inferred from the quoted accuracy of the
distance measurement device; and a gross error in one of the distance mea-
surements (or an error in the quoted position of one of the known points)
will still give a seemingly plausible solution for the new point’s position.
To overcome both of these problems, a fundamental principle of survey-
ing is to take redundant readings: that is, to take more measurements than
are strictly necessary to fix the unknown quantities. Any large inconsis-
tency in the readings will then indicate a gross error in the measurements
or the data, while any small inconsistencies will give an unbiased indication
of the likely accuracy to which the point has been fixed.
When several new points are to be fixed simultaneously, it can become
quite difficult to ensure by simple inspection that enough suitable readings
have been taken or planned to ensure redundancy throughout the network.
This soon becomes apparent, though, when the readings are adjusted by
computer (see Section 2.4 below). For this reason, many adjustment pro-
grams include a planning mode, which enables a proposed scheme of obser-
vations to be validated for redundancy before it is carried out. A surveyor
is strongly advised to carry out such a check, if there is any doubt about the
redundancy of a proposed scheme of observations.

2.3 STIFFNESS

In addition to being redundant, a network (and its associated observations)

should also be ‘stiff’—in other words, the relative positions of control
points and the scheme of observations should be arranged such that any
significant movement of one of the points would cause a correspondingly
significant change in at least one of the observations. This ensures that the
positions of unknown points are established to the highest possible accu-
racy, using the instruments which are available.
There is an exact analogy (as with redundancy) between a ‘stiff’ network
and a stiff structure. The pin-jointed structure shown in Figure 2.1(a) is stiff,
because any given deflection of point C requires that member AC or BC (or
both) must lengthen or shorten by a similar amount. In Figure 2.1(b), by
 General Principles of Surveying 11

A
A
C C

(a) (b)

Figure 2.1 Stiff and non-stiff structural frameworks.

contrast, the structure is much less stiff since C can make quite large vertical
movements with relatively small changes in the lengths of the two members.
The corresponding situation in surveying is shown in Figure 2.2, where
points A and B are known points, and C is unknown; and the distances AC
and BC have been measured. As with the structure, Figure 2.2(a) shows a
stiff network, in which any significant movement of point C would involve
equally significant changes to one or both of the measured distances;
whereas in Figure 2.2(b), C could move significantly in the north/south
direction without greatly affecting either of the distances.
If angle measurements are used as well, this corresponds to adding gus-
set plates to the structure, which increases its stiffness by removing the
freedom in the pin joints.
As with redundancy, it can be quite difficult to determine by inspection
whether a proposed scheme of observations will result in a stiff network.
Again, though, an adjustment program with a ‘planning’ facility will pro-
vide a good prediction of how accurately the unknown points will be fixed,
if the likely accuracy of the planned observations is known.

North
A North

A
C C
B
East
East = Known point
B
= Measured distance
(a) (b)

Figure 2.2 Stiff and non-stiff survey networks.

12 Plane and Geodetic Surveying

2.4 ADJUSTMENT

As explained in Section 2.2, the position of new points should always be

found by taking more observations than are strictly necessary. Inevitably,
then, the resulting readings will be in conflict; because of the small random
errors in the readings, there will be no single set of positions for the new
points which will be in exact agreement with all the measurements.
To resolve this problem, some form of ‘adjustment’ is usually applied to
the calculated position of the point, to give the best fit with the measurement
data. The commonest method is called least-squares adjustment, which
chooses positions for the new points such that the sum of the squares of the
residual errors* is minimised. This gives the most likely positions for the
new points, assuming that the observation errors are normally distributed.
A good understanding of what adjustment can, and cannot, achieve is
important for a surveyor. Essentially, it is a statistical process which gives
the most likely position for each new point, assuming that the observation
errors are random and normally distributed. If this is not the case, the
results may be misleading or inaccurate. In particular, least-squares adjust-
ment will give a false impression of accuracy if there are systematic errors
present in the data, e.g., if all distance measurements are made using a
device which is poorly calibrated. It will also generate misleading results if
the user is tempted to reject any seemingly ‘bad’ observations purely on the
grounds that they do not appear to agree well with the others.
Adjustment is described in greater detail in Chapter 11.

2.5 PLANNING AND RECORD KEEPING

A successful survey requires an appropriate set of measurements to be

taken and recorded without unnecessary deployment of human resources
or equipment. This can only be achieved by means of planning. The follow-
ing guidelines will improve the quality of any surveying work.

1. Establish clearly what the purpose of the survey is and what addi-
tional uses it might be put to in the future. This will determine the
number and locations of control points and the accuracy to which
their positions must be found.
2. Find a suitable map or satellite photograph of the site to be surveyed.
This will help in the creation of a possible network of control points,
in suitable locations and with adequate stiffness. It will also show the
* The residual error is defined as the difference between an observed angle or distance, and
the calculated value based on the assumed position(s) of the new point(s).
 General Principles of Surveying 13

approximate scale of the work and will help in detecting gross errors
in angle and distance measurements.
3. Visit the site if at all possible. Check whether control stations can
be sited at the places indicated by Step 2, and make a note of what
will be needed to build them. If conventional instruments are to be
used, check whether the necessary lines of sight exist between the sta-
tion locations, using ranging rods if necessary. If GNSS is to be used,
check that the relevant stations have a clear view of the sky. Make
notes of any features on the site (cliffs, ditches, etc.) which might
make it difficult to move from one station to another.
  A few simple instruments may also help at this stage. A compass
can be used to estimate horizontal angles, and a clinometer will mea-
sure approximate vertical angles. A hand-held GNSS receiver will give
the approximate co-ordinates of points and estimates of the distances
between them. If this is not possible, a hand-held laser measure can
be used, or the distances can be paced.
4. Plan a set of observations which will establish the control network to
the required accuracy at minimum cost. This is generally best done
by working ‘from the whole to the part’: accumulated errors are min-
imised by first forming an accurate framework covering the whole
area, and then adding further control stations to whatever accuracy is
necessary. Accurate measurements require expensive equipment and
longer observation times, so this type of consistent approach will give
the most economical result.
  The planning function in an adjustment program is very useful here.
The eventual quality of a network can be reliably predicted by enter-
ing approximate observations (such as the compass angles above),
together with estimates of the accuracy to which the final measure-
ment will be made.* Different observations can then be included in
the scheme, to see which combination will give an adequate accuracy
for minimum investment. Make sure, though, that there are enough
observations so that one or more could be rejected without unaccept-
able loss of accuracy or redundancy. The time spent travelling to and
from a site is usually much greater than that needed to take a few
‘spare’ measurements while an instrument is set up.
5. Plan the fieldwork in detail to make sure that all the necessary mea-
surements are taken with the minimum deployment of people and
equipment. Each member of the team should know who will take
which measurements, at which locations, and with what instruments.
* The approximate observations establish the geometry of a network to sufficient accuracy
for its eventual stiffness to be determined. This, combined with the accuracy of the final
observations, determines the accuracy to which the points in the network will eventually
be fixed.
14 Plane and Geodetic Surveying

6. If possible, arrange that all fieldwork has redundancy, and that the
computations are carried out such that no gross error (Section 2.1.1)
will pass undetected. If some of the error checks can be carried out in
the field while the equipment is still set up on station, then the cost of
correcting any error will be greatly reduced.
7. Before leaving base, make sure that all batteries are fully charged and
that any necessary co-ordinate data, transformations, etc., have been
downloaded into those instruments that need them. Make sure that
everyone is familiar with the instruments they will be using: get unfa-
miliar instruments out, read the instruction manuals, and practise
their use.
8. Ensure that each group of surveyors keeps a diary of what is done,
including a summary of the weather, on each day. If an error is discov-
ered later, a good diary can be invaluable in pinpointing the source of
the problem, and thus showing which measurements may need to be
repeated.
9. Make sure that observation records are complete, and will not degrade
with time—the data generated during a surveying job may need to
be consulted many years after it was initially made. Observations
recorded on paper should be checked for legibility and completeness,
and stored in a dry condition; electronic data should be stored on a
long-term medium, such as a CD-ROM. For important jobs, copies
of the data should be made and stored in a different location from the
originals—the cost of this is minuscule compared to the cost of tak-
ing the measurements again. Finally, a brief summary of the data will
greatly assist any subsequent attempt to re-inspect some part of it.