THEORY OF ERRORS

 Measurements/observations/parameters/quantities
 In geospatial engineering, both the term measurements and observations are used
to mean the same thing usually referring to the process of the acquisition of data or
measured quantities.
 All observations/measurements/measured quantities recorded are usually not ready for
use without processing depending on precision, accuracy and degree of reliability. Mostly
we measure/observe many times, by different methods, using different instruments and
people which are then combined to determine a quantity as well as the quality of the
measurements
 observations/measurements/measured quantities must therefore be before being used to
compute the required parameters/computed quantities
1
 THEORY OF ERRORS

Sources of errors
Measurement errors stem from four main sources, namely:
 1. Instrumental errors. These errors are caused by imperfections in instrument
construction or adjustment or malfunctioning instruments.
 2. Natural errors. These errors are caused by changing conditions in the surrounding
environment, including variations in atmospheric pressure, temperature, wind,
gravitational fields, and magnetic fields i.e. unfavorable physical environment
 3. Human/Personal errors. These errors arise due to limitations in human senses, such as
the ability to read a micrometer or to center a level bubble.
 4. Improper procedures. These errors arise due to using wrong measurement methods
2
 THEORY OF ERRORS

 Types of errors
 Normally, one may distinguish three types of errors: systematic errors, gross errors and
random errors.
 Gross errors are also called blunders or outliers and are large errors in measurements
mainly due to human mistakes, malfunctioning instruments or wrong measurement
methods.
 In principle, gross errors are not permitted and should be avoided by surveyors’
carefulness and control routines.
 For example, it can happen that a surveyor might write 500 32’ 50:9” in his field
observation protocol when the actual reading on the theodolite is 500 32’ 5:9”
3
 THEORY OF ERRORS
 Types of errors
 Systematic errors are errors which follow certain physical or mathematical rules and
often affect surveying results systematically. The causes of this kind of errors can be the
instruments used, physical environment in which measurements are made, human factors and
measurement routines.
 One example of systematic errors is the constant error of -5 cm for a tape. This constant
error will cause a systematic error to all distance measurements by this tape
 To avoid or reduce systematic errors, one may
 (a) calibrate carefully instruments before field work starts;
 (b) design and use suitable measurement routines and procedures which can reduce or
eliminate possible systematic errors;
 (c) if possible, correct measurement results afterwards. 4
 THEORY OF ERRORS
 Types of errors
 Random errors or stochastic errors are the errors which remain after systematic and gross
errors have been eliminated and are assumed to be randomly distributed. Random errors can be
caused by human factors, instrument errors, physical environment and measurement routines
 The characteristics random errors
 Random errors are believed to follow normal distribution i.e.
 The arithmetic mean of 𝑖 approaches zero when the number 𝑛 of observations approaches infinity:

σ𝑛
𝑖=1 𝜀𝑖
 lim =0
𝑛→∞ 𝑛

 Positive errors and negative errors with same magnitude occur roughly at equal frequency;
 Errors of smaller magnitude occur more often ( at a higher frequency) than errors of larger magnitude;
5
 THEORY OF ERRORS
 Quality of measurements/Observations
 Several terms are used to define or describe the quality of measurements
 True Value of a Quantity (𝝉)
 The value of a quantity which is absolutely free from any error is called the true value.
 It can never be found out and the true value of a quantity is indeterminate
 Most Probable Value of a Quantity (MPV)
 The most probable value of a quantity is one which is most likely to be true value than any other values.
 This is most likely to be having minimum errors or has the highest probability of occurrence
 In case of direct observations of equal weight, the most probable value is the arithmetic mean.
 In case of direct observations of unequal weights, the most probable value is the weights; the most probable
value is the weighted arithmetic mean
6
 THEORY OF ERRORS
 Quality of measurements/Observations
 Precision
 Precision describes the degree of repeatability of the
observations and is an internal measure of random errors and
their effects on the quality of observation. That is how close
measured or computed values are to each other.
 Precision is the degree of consistency between observations
based on the sizes of the discrepancies in a data set
 Accuracy
 Accuracy is the measure of the absolute nearness of a measured
quantity to its true value. Since the true value of a quantity can
never be determined, accuracy is always an unknown
7
 THEORY OF ERRORS
 Quality of measurements/Observations
 Reliability
 Reliability concerns the capability of observations against gross and systematic errors. Observations or surveying methods
with high reliability can easily detect gross and systematic errors, and are said to be robust
 Measurement/Observation error (e)
 If 𝝉 is the true value of a quantity and 𝒍 is the observed/measured value then the error 𝒆 in measurement is given by:
 𝒆= 𝝉−𝒍
 Since we don’t know the true value 𝝉, we use adjustment techniques to estimate the value as the most probable value
usually represented by 𝑴𝑷𝑽. The difference between the most probable value and the measured quantity is referred to as
the residual and is given by:
 𝒆𝒊 = 𝝁 − 𝒍𝒊 = 𝑴𝑷𝑽 − 𝒍𝒊
 This variation in measurements (residual) is taken as the measurement error. For measurements taken using the same
instrument, at the same location, using the same method and by the same observer, the MPV is usually taken to be the
mean of the measurements. 8
 THEORY OF ERRORS
 Quality of measurements/Observations
 Average error (Mean error) (ത
𝒆)
 Most measurement errors are random errors that behave in a random way.
 Therefore, in practice it is very difficult or impossible to describe each individual error which occurs under a specific
measurement condition (i.e. under a specific physical environment, at a specific time, with a specific surveyor, using a
specific instrument, etc).
 However, for simplicity we often prefer to use one or several simple index numbers to judge the quality of the obtained
measurements (how good or how bad the measurements are).
σ 𝝁−𝒍𝒊
 This is the mean of all measurement errors, taken without regard to the sign and is given as: 𝒆ത = 𝒏

 Standard deviation (𝝈) (Also called Root-Mean-Square-Error) (R.M.S.E)

 Standard deviations are index numbers used to describe the precision of the measurements and the derived quantities
and is the widest used in statistics. It is given as:

σ(𝑴𝑷𝑽−𝒍𝒊 )2 σ 𝒆𝟐𝒊
 𝝈= =
𝒏−𝟏 𝒏−𝟏 9
 THEORY OF ERRORS
 Quality of measurements/Observations
 Example1:
 A distance between two points was measured ten times as follows
 234.025, 234.020, 234.018, 233.999, 234.016, 234.021, 234.012, 234.007, 234.005, 234.016

𝒍𝒊 𝐞𝒊 =𝐌𝐏𝐕-𝒍𝒊 (𝑴𝑷𝑽 − 𝒍𝒊 )𝟐 = 𝐞𝟐𝒊

1 234.025 0.011 0.000121
2 234.020 0.006 0.000036
3 234.018 0.004 0.000016
4 233.999 -0.015 0.000225 σ 𝒆𝟐𝒊 0.000589
5 234.016 0.002 0.000004
𝝈= = =0.008
𝒏−𝟏 𝟗
6 234.021 0.007 0.000049
7 234.012 -0.002 0.000004
8 234.007 -0.007 0.000049  Therefore the distance between
9 234.005 -0.009 0.000081 the two points is 234.014 + 0.008
10 234.016 0.002 0.000004
MPV(𝝁)=234.014 σ 𝝑𝟐𝒊 =0.000589 10
 THEORY OF ERRORS

 Quality of measurements/Observations
 Example 2:
 A line PQ is measured severally in three sections 𝑎, 𝑏 𝑎𝑛𝑑 𝑐 as shown below:
 a: 90.21, 90.17, 90.18, 90.20, 90.22, 90.21
 b: 45.59, 45.61, 45.63, 45.57, 45.60, 45.60, 45.62, 45.64, 45.61
 c: 25.85, 25.83, 25.87, 25.86, 25.90, 25.82, 25.84, 25.86, 25.88, 25.87
 Determine the length of PQ and its uncertainty/standard deviation

11
 THEORY OF ERRORS
 Solution Example 2:
Section c
Section a Section b
𝒍𝒊 𝝑𝒊 =𝒆𝒊 =𝝁-𝒍𝒊 𝝑𝟐𝒊 =𝐞𝟐𝒊 𝒍𝒊 𝒆𝒊 = 𝝁 - 𝝑𝟐𝒊 =𝐞𝟐𝒊 𝒍𝒊 𝒆𝒊 = 𝝁 - 𝝑𝟐𝒊 =𝐞𝟐𝒊
1 90.21 0.012 0.000144 𝒍𝒊 𝒍𝒊
2 90.17 -0.028 0.000784 1 45.59 -0.02 0.0003 1 25.85 -0.008 0.000064
3 90.18 -0.018 0.000324 2 45.61 0.00 0.0000 2 25.83 -0.028 0.000784
4 90.20 0.002 0.000004 3 45.63 0.02 0.0005 3 25.87 0.012 0.000144
5 90.22 0.022 0.000484 4 45.57 -0.04 0.0014 4 25.86 0.002 0.000004
6 90.21 0.012 0.000144 5 45.60 -0.01 0.0001 5 25.90 0.042 0.001764
MPV(𝝁)=90.2 σ 𝝑𝟐𝒊 =001884 6 45.60 -0.01 0.0001 6 25.82 -0.038 0.001444
7 45.62 0.01 0.0001 7 25.84 -0.018 0.000324
8 45.64 0.03 0.0010 8 25.86 0.002 0.000004
9 45.61 0.00 0.0000 9 25.88 0.022 0.000484
σ 𝒆𝟐𝒊 0.001884
𝝈= = = ±𝟎. 𝟎𝟏𝟖 𝝁=45.608 σ 𝝑𝟐𝒊 =0.0036 10
𝒏−𝟏 𝟓 25.87 0.012 0.000144
𝝁=25.858 σ 𝝑𝟐𝒊 =0.0052
σ 𝒆𝟐𝒊 0.0036
𝝈= = = ±𝟎. 𝟎𝟐𝟏
𝒏−𝟏 𝟖 σ 𝒆𝟐𝒊 0.0052 12
𝝈= = = ±𝟎. 𝟎𝟐𝟒
𝒏−𝟏 𝟗
 Hence the length of PQ is = 𝟗𝟎. 𝟐 + 𝟒𝟓. 𝟔𝟏 + 𝟐𝟓. 𝟖𝟔 = 𝟐𝟔𝟏. 𝟔𝟕𝒎
 THEORY OF ERRORS
 Definition of Error propagation
 Error Propagation is defined as the effects errors in measurements/observations on the computed
quantities or parameters and also the effects of errors in the computed quantities or parameters on the
subsequently computed quantities
 It is how the standard errors accumulate when you use the measurements to compute/derive other
quantities.
 Error propagation can also be from one measurement to another e.g. Random errors in observations
such as instrument centering errors, reading errors, and pointing errors propagate into the observed
angles, distances, and elevation differences
 It is a calculus derived statistical calculation designed to combine uncertainties/standard errors from
multiple measured/computed variables, in order to provide the uncertainty of the computed or derived
parameters.
 Error propagation is dependent on how the measurements/quantities are manipulated to obtain the
computed/derived quantities or parameters e.g. addition, subtraction, multiplication, division etc.
13
 THEORY OF ERRORS
 Error propagation
 Addition/Subtraction of measured quantities
 If you have a computed quantity R from the measured values (measurements) X, Y, and Z, with
uncertainties/standard errors 𝜎𝑥 , 𝜎𝑦 , 𝑎𝑛𝑑 𝜎𝑧 , such that the result, R, is the sum or difference of these
quantities/measurements given as
𝑅 =𝑋+𝑌−𝑍
 Then the uncertainty or standard error in R, 𝜎𝑅 is given by:

𝜎𝑅 = 𝜎𝑋2 + 𝜎𝑌2 + 𝜎𝑍2

 Please note that the rule is the same for addition and subtraction of quantities.
 In Example 2 above:
 We had the length of line PQ as = 𝟗𝟎. 𝟐 + 𝟒𝟓. 𝟔𝟏 + 𝟐𝟓. 𝟖𝟔 = 𝟐𝟔𝟏. 𝟔𝟕𝒎
 The standard error of the line would then be:

𝜎𝑃𝑄 = 𝜎𝑋2 + 𝜎𝑌2 + 𝜎𝑍2 = 𝜎𝑎2 + 𝜎𝑏2 + 𝜎𝑐2 = 0.0182 + 0.0212 + 0.0242 = ±0.037
 THEORY OF ERRORS
 Error propagation
 Addition/Subtraction of measured quantities
 Example 3:
Given the x-coordinates of two points 1 & 2 were measured as follows
𝑋1 = 1009.3 ± 0.2𝑚 and 𝑋2 = 1114.3 ± 0.3
Determine the difference in the x-coordinates (∆𝑋) and its error.
The difference in coordinates is given as
∆𝑋 = 𝑋2 − 𝑋1 = 1114.3 − 1009.3 = 105𝑚
The error in the difference in coordinates is given as

𝜎𝑅 = 𝜎𝑋2 + 𝜎𝑌2 + 𝜎𝑍2

∴ 𝜎∆𝑋 = 𝜎𝑋22 + 𝜎𝑋21 = 0.32 + 0.22 = 0.09 + 0.04 = 0.13 = ±0.36 𝑚 15

Hence we have ∆𝑋 = 105 ± 0.36𝑚

 THEORY OF ERRORS

 Error propagation
 Multiplication/Division of measured quantities
 If you have a computed quantity R from the measured values (measurements) X, Y,
and Z, with uncertainties/standard errors 𝜎𝑥 , 𝜎𝑦 , 𝑎𝑛𝑑 𝜎𝑧 , such that the result, R, is
the multiplication and/or division of these quantities/ measurements given as:
𝑋. 𝑌
𝑅=
𝑍
 Then the uncertainty or standard error in R, 𝜎𝑅 is given by:

𝜎𝑋 2 𝜎𝑌 2 𝜎𝑍 2
𝜎𝑅 = 𝑅 + +
𝑋 𝑌 𝑍

 Please note that for the purpose of error calculation there is no difference between
multiplication and division. 16
 THEORY OF ERRORS

 Error propagation
 Multiplication/Division of measured quantities
 Example 4:
The two sides of a rectangular piece of land have been measured as 𝑠1 = 100𝑚 and 𝑠2 = 50𝑚 with standard
deviations as 𝜎𝑠1 = ±0.3 and 𝜎𝑠2 = ±0.2 respectively. Determine the area of the parcel of land together
with its standard deviation/uncertainty.
 The area of the parcel is given as 𝐴𝑟𝑒𝑎(𝐴) = 𝑠1 × 𝑠2 = 100𝑚 × 50𝑚 = 5000𝑚2
 The error in the area is given as

𝜎𝑋 2 𝜎𝑌 2 𝜎𝑍 2
𝜎𝑅 = 𝑅 + +
𝑋 𝑌 𝑍
2 2 2 2
𝜎𝑠1 𝜎𝑠2 0.3 0.2
∴ 𝜎𝐴𝑟𝑒𝑎 = 𝐴𝑟𝑒𝑎 + = 5000 +
𝑠1 𝑠2 100 50
17
𝜎𝐴𝑟𝑒𝑎 == 5000 0.000025 = 5000 × 0.05 = ±25
𝐴𝑟𝑒𝑎 = 5000 ± 25𝑚2
 THEORY OF ERRORS

 Error propagation
 Multiplication with a constant
 What if you have measured the uncertainty in an observable X, and you need to multiply it with a constant that
is known exactly? What is the error then? This is easy: just multiply the error in X with the absolute value of the
constant, and this will give you the error in R:
𝑅 = 𝐶. 𝑋
𝜎𝑅 = 𝐶 . 𝜎𝑋
 Example 5:
Compute the circumference of a circle with its error if the diameter was measured as 𝐷 = 4.5 ± 0.5𝑚.
 The circumference of circle (C) = 𝜋𝐷 = 3.14159 × 4.5 = 3.14159 × 4.5 = 14.137𝑚
 The error in the circumference of circle computed is given as
𝐸𝑟𝑟𝑜𝑟 𝑖𝑛 circumference = 𝜋 𝜎𝐷 = 3.14159 × 0.5 = 3.14159 × 0.5 = ±1.57
Therefore the circumference is = 14.137 ± 1.57𝑚 18
 THEORY OF ERRORS

 General function
 And finally, we can express the uncertainty in R as a general functions of one or more observables.
 If R is a function of measurements (X.Y,Z…), written as
𝑅 = 𝐹 𝑋, 𝑌, 𝑍 … . .
 Then the uncertainty in R is obtained by taking the partial derivatives of R with respect to each measurement such that
2 2 2
𝜕𝑅 𝜕𝑅 𝜕𝑅
𝜎𝑅 = .𝜎 + .𝜎 + .𝜎 +⋯
𝜕𝑋 𝑋 𝜕𝑌 𝑌 𝜕𝑍 𝑍
 All rules that we have stated above are actually special cases of this last rule. For example we had: 𝑅 = 𝑋 + 𝑌 − 𝑍 then
2 2 2
𝜕𝑅 𝜕𝑅 𝜕𝑅 2 2 2
𝜎𝑅 = .𝜎 + .𝜎 + .𝜎 = 1. 𝜎𝑋 + 1. 𝜎𝑌 + −1. 𝜎𝑍
𝜕𝑋 𝑋 𝜕𝑌 𝑌 𝜕𝑍 𝑍
 Which reduces to:
𝜎𝑅 = 𝜎𝑋 2 + 𝜎𝑌 2 + 𝜎𝑍 2

 If we have: 𝑅 = 𝑋𝑌, then

2 2
𝜕𝑅 𝜕𝑅 2 2 19
𝜎𝑅 = .𝜎 + .𝜎 = 𝑌. 𝜎𝑋 + 𝑋. 𝜎𝑌
𝜕𝑋 𝑋 𝜕𝑌 𝑌
 THEORY OF ERRORS
 Error propagation
 General function
 Example 6:
The two sides of a rectangular piece of land have been measured as 𝑠1 = 100𝑚 and 𝑠2 = 50𝑚 with standard
deviations as 𝜎𝑠1 = ±0.3 and 𝜎𝑠2 = ±0.2 respectively. Determine the area of the parcel of land together with its
standard deviation.
The area of the parcel is given as
𝐴𝑟𝑒𝑎 = 𝑠1 × 𝑠2 = 100 × 50 = 5000𝑚2
 The error in the area is given as
2 2 2 2
𝜕𝑅 𝜕𝑅 𝜕𝐴𝑟𝑒𝑎 𝜕𝐴𝑟𝑒𝑎 2 2
𝜎𝑅 = .𝜎 + .𝜎 = . 𝜎𝑠1 + . 𝜎𝑠2 = 𝑠2 . 𝜎𝑠1 + 𝑠1 . 𝜎𝑠2
𝜕𝑋 𝑋 𝜕𝑌 𝑌 𝜕𝑠1 𝜕𝑠2

𝜎𝐴𝑟𝑒𝑎 = 50. 𝑋0.3 2 + 100𝑋0.2 2 = 15 2 + 20 2 = 625

𝜎𝐴𝑟𝑒𝑎 = ±25
20

∴ 𝐴𝑟𝑒𝑎 = 5000 ± 25𝑚2