Visual Image

UNIT 9 GROUND TRUTH DATA Interpretation: Selected

Examples
COLLECTION

Structure
9.1 Introduction
Objectives
9.2 Ground Truth Data
Definition
Need and Purpose of Data Collection
9.3 Planning and Collection of Ground Truth Data
What to Collect?
When to Collect?
Where to Collect?
How to Reach to the Ground Truth Sites and Locate Yourself on the Ground?
How to Collect?
What to Carry While Going for Data Collection?
Estimating Cost of Data Collection
Do’s and Don’ts
9.4 What to Do Once the Data is Collected?
9.5 Activity
9.6 Summary
9.7 Unit End Questions
9.8 Reference
9.9 Further/Suggested Reading
9.10 Answers

9.1 INTRODUCTION
In the previous two units, you have studied the concepts of visual image
interpretation and importance of image interpretation elements and also how
to interpret images. You can interpret an image if you are familiar with the
area it represents. Now, imagine a situation when you are not familiar with the
study area then in that case how would you interpret the image of that area? If
you have somehow interpreted the image, how would you ensure that the
information derived from that image is correct and it truly represents the
features as they are on the ground? To relate the image and image derived
information with the real features on ground you need to collect some data on
ground or in situ. Ground data collection helps in building the linkage between
the image and ground reality.

In this unit, you will study about ground truth (field) data, its importance and
planning and different methods of data collection.

45
Image Interpretation Objectives
After studying this unit, you should be able to:
• define ground truthing;
• discuss importance of ground truth data collection;
• explain the methods of sampling;
• describe how, when and where to collect ground truth data;
• list out the things you need to carry while going for ground truth data
collection; and
• point out the things to be remembered during field campaigns.

9.2 GROUND TRUTH DATA

In this section, we would first define ground truth and then discuss about the
need and purpose of collecting ground truth data.

9.2.1 Definition
Ground truth is a term used in a range of remote sensing techniques. It
generally refers to the data collected ‘on site’, on surface of the Earth
regarding characteristics of the Earth surface features. Ground truth data are
complementary to the remote sensing data as it helps to link the image data to
the ground reality.

You should be aware that though the term ‘ground truth’ contains the word
‘truth’, the ground truth data is not always true as it may have errors in it due
to factors such as:
• errors introduced during data collection
• data processing and
• deriving inferences from them.
The process of ground truth data collection involves collection of information
about the area of study from field survey, analysis of aerial photographs or
high spatial resolution data and personal experience. Ground truth is usually
done on field/site, performing surface observations and measurements of
various properties of the ground features that are being studied on the remote
sensing images. It also involves taking geographic coordinates of the sampling
locations and comparing them with the coordinates of the image features to
understand and analyse the location errors and its affect on your study and
analysis outputs.

More specifically, ground truthing refers to a process in which a feature on a

remote sensing image is compared to the ground reality in order to verify the
image features. In cases, where there are maps as output of remote sensing
data interpretation and analysis, it allows to help in determining accuracy of
the maps generated by the remote sensing data analysts or image analysis
software. This minimise errors in the computer assisted thematic information
extraction procedures.

46
The process of collection of ground truth data is known as ground truthing. Ground Truth Data
Collection
Some people also use the terms field work or field survey or ground survey or
surface data collection or reference data collection. Although, the term
‘ground truth’ is more widely used, it is not an appropriate term because data
collection may not be only from ground but also from air, water or ice. Then,
in those cases observations made from air should be called as air truth and
similarly if the observations are made from sea it should be called as sea
truth. Hence, some investigators suggest that the term ground truth may be
abandoned in favour of the term in situ data. Some people prefer to use the
term surface truth. Another suggested term is reference data, which appears to
be more appropriate, however, throughout this unit, the terms ‘ground truth’ or
‘ground truth data’ are used as these are more commonly used terms in
geoinformatics related literature.

9.2.2 Need and Purpose of Data Collection

Collection of ground truth data is an essential part in geoinformatics. Let us
now see what the actual need of the ground truthing is. We would also briefly
discuss about a situation where you may not need to collect any ground truth
data for your study.

As you know different surface features such as water, soil, vegetation, etc.
have their characteristic spectral signatures. Hence, in an ideal remote sensing
world, nature of the Earth surface features should be completely specified by
the spectral signatures. However, in the real world, the complexity of Earth
surface features, effects of atmosphere and indistinctness of spectral signatures
act as limiting factors to remote sensing without ground truth data. Ground
truth data collection is complementary to the synoptic overview provided by
satellites as it helps to link the image data to the context of Earth surface
features existing on the ground. So, ground truth is important for image
interpretation.

Irrespective of the method of interpretation, generally, the ground truth data is

divided into two parts. Ground truth data belonging to the first part is used as
a guide (training sites) for image interpretation and ground truth data
belonging to the second part is used for validating information (accuracy
assessment sites) derived from image data. When the identity and location of Decision rule is a function
image features of interest are known through a combination of ground truth which designates a specific
data, available maps and personal experience, those areas are known as condition or combination
training sites. The spectral characteristics or signatures of these areas are used of conditions in the
to train remote sensing data processing software for classifying rest of the decision-making process
and to take appropriate
image using a decision rule. There are several decision rules about which you action if those conditions
would be studying in detail in Unit13 of MGY-002. Ground truth sites exist.
belonging to the second part are used by analyst to validate maps.

Let us now summaries here the purpose of ground truth data collection. It is
carried out typically for the following purposes:
• to calibrate a remote sensor
• to help in remote sensing data correction
• to provide reliable data to identify each feature of interest present in
image to aid and guide the process of image interpretation/analysis
47
Image Interpretation • to locate representative areas of each image feature in order to generate
their spectral signatures so as to model the spectral behaviour of specific
Earth surface features and
• to assess the accuracy of thematic maps generated and information
derived from images and validate them.
Spend Check Your Progress I
5 mins
1) List out the purpose of ground truth data collection.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................

9.3 PLANNING AND COLLECTION OF GROUND

TRUTH DATA
In the previous section, we have discussed that the practice of remote sensing
depends largely upon the use of accurate ground observations. However,
collection of ground data remains one of the least systematised aspects of the
field. The errors may result in inaccurate interpretation of remote sensing data.
Hence, it is critical to carefully plan for ground truthing and collect ground
truth data that best represents the ground reality. You shall study here the
process of ground truth data collection and methods of data collection, which
would guide you to properly plan for ground truth data collection for your
study.

The primary objectives of any remote sensing based study include one or more
of the followings:
• to produce a map of surface features of a specific theme
• to validate or evaluate the final product such as maps
• to model spectral response of a surface material and its biophysical
characteristics
• to develop or test image data processing techniques and
• to calibrate remote sensing sensors with the recorded spectral response of
surface features.
Outcome of ground truthing may vary depending upon the project objectives.
The outcomes may be simply an aerial photograph or high resolution remote
sensing data annotated with current ground cover types or a set of ground truth
data sheets containing information on measurements and observations for each
sample site.

While planning for ground truth data collection due considerations should be
given to the following points:

48
• project objectives Ground Truth Data
Collection
• nature of features being studied
• data collected should represent all features present in the study area
• data should represent environmental conditions
• selection of appropriate sampling strategy and
• costs and time of data collection.

9.3.1 What to Collect?

What kind of data should you collect from the ground would depend upon the Ground truth data
project objectives. You should begin by examining project objectives and then collection plan should
decide upon what kind of data you should collect to fulfill those objectives be compatible with the
and accordingly set objectives of ground measurement. Based on the nature of resolution of the sensor,
observations made and recorded ground truth data can be grouped under spatial scale of the
qualitative and quantitative data. Qualitative information includes name of the landscape and kinds of
class, its location, etc. whereas quantitative information require some analyses to be carried
measurement to be made on the ground. out.

Ground truth data includes minimum three elements i.e. attribute, location
and time. Attribute information are the observations which describe ground
conditions at a given ground truthing site. Locational information is helpful in
linking the ground observations to match with corresponding points in images
and observations related to time and date. Data should also have records about
identity(/ies) of the person(s) who are involved in the collection process,
calibration information for instruments, weather conditions and other
components as required in specific projects.

Ground truth data may come from a variety of sources such as field surveys or
measurements, interpretation of very high resolution images or aerial
photographs and library records/reports. It can be categorised into following
two types:
• thematic data and
• spectral data

Thematic data provides information about distribution of features related to a

specific theme such as water bodies, land use/cover features, crops, landforms,
rock types. Spectral data is not generally available in existing records/reports
and are to be collected from field using some instruments such as radiometer,
goniometer.

According to Campbell (1996), ground truth data can also be grouped under James B. Campbell is a
the following four types: Professor in Geography at
Virginia Tech in
i) Nominal Data Blacksburg, USA, where
he teaches remote sensing
• consists of qualitative designations applied to regions delineated on and geomorphology. He
imagery that convey basic differences from adjacent regions such as has worked in the fields of
forest, crop land, urban land, river and lake forestry, geology,
agronomy and environ-
• nominal labels originate from several alternative sources such as mental sciences.
from the classification system
49
Image Interpretation • is easy to collect at points/small areas and
• timing of field observation should match those of images because
physical characteristics of some of the classes may vary with time.

ii) Biophysical Data

• consists of measurements of physical characteristics collected, which
describes the type, size, form and spacing of plants forming
vegetative cover or texture and mineralogy of soil surface
• when associated with nominal data, they document the meaning of
nominal data
• typical data might include leaf area index, biomass, soil texture, soil
moisture, net primary productivity and so on and
• typically apply to points, so must be linked to areas by averaging
values from several observations within an area.

iii) Field Radiometry

• it is radiometric data collected with the help of hand held sensors,
which permits analyst to relate brightness as recorded by aerial or
space based sensors
• hand held device can be mounted on some platforms to simulate
viewing perspective of airborne or satellite sensors and
measurements are coordinated with time of acquisition of aircraft or
satellite data
• hand held device is attached to a notebook computer to record the
spectra
• brightness can be recorded as radiance or reflectance and
• it is generally used for sensor calibration purpose.

iv) Locational Information

• consists of information regarding locations/position (longitude and
latitude), which is generally collected using GPS
• data can be in point, line or polygon form
• permits attributes gathered in the field to match with imagery of the
same region
• locations of distinctive landmarks, visible both in field and on
image, permit the analyst to match the information and
• accuracy of location is important.

9.3.2 When to Collect?

Ideally, ground truth data should be collected at the time of image acquisition.
Ground truthing should be carried out at previously determined locations for
the features of interest along with other features present in the image so as to
truly represent ground reality at the time of image acquisition. Permanent
50
features can be observed anytime before or after the time of image acquisition. Ground Truth Data
Collection
However, due care should be given to the fact that ground truthing should be
done before characteristics of the features of interest change. This is more
important in the cases where features of interest are highly dynamic such as
crops and vegetation in the season of their growth, soil moisture, tidal
conditions, etc.
Another question of timing of ground data collection is concerned with
whether to collect ground data before or after the interpretation. If you do not
have any idea about characteristics of ground features you should first collect
ground data and then carry out interpretation task. However, if you have some
idea about spectral characteristics of features of interest then you may like to
first identify and map major classes on the image based on spectral variations
in the image and then carry out ground data collection.

9.3.3 Where to Collect?

Ideally, ground truth data should be collected in the entire study area.
However, this is possible only if the study area is small, accessible and also
there are sufficient resources (time, budget, personnel, etc.) available.
However, when the area of study is very large due to practical reasons such as
inaccessibility of the study area, time and budget constraints, it is not feasible
to conduct ground truthing in the entire study area. Hence, ground truthing is
generally carried out in some selected sites, which are representative of the
entire study area.
For convenience, we will discuss this under the following headings:
• number of observations to be made
• considerations while planning sampling patterns and
• types of sampling patterns.
Number of Observations to be Made
Before you decide where to collect data, you should first decide on the number
of observations to be made. It is fundamental and complex task. You should
keep in mind the following points while determining the number of
observations:
• areal extent of the study area
• number of features present in image and their variations
• number of observations required for each category and
• number of observations required to have confidence and accuracy of map
products.
Considerations While Planning Sampling Patterns
Sampling pattern determines how observations would be arranged on ground. Sample is a part or piece
that is drawn from and
It governs amount of time that would be required for collecting data and also analysed to estimate the
accuracy of the output. While planning the sampling strategy, following characteristics of the
factors should be considered: whole.

• objectives of the study

• accessibility of the study area
51
Image Interpretation • nature and characteristics of the features being studied
• time constraint
• number of observations to be made and
• availability of other resources.
Types of Sampling Patterns
Commonly used sampling patterns are the following:
• Simple Random Sampling Pattern: In this pattern, observation sites are
randomly chosen (Fig. 9.1a). Randomness ensures that all parts of the
study area would have an equal chance of being sampled without any
human bias. Observation sites are selected by dividing the study area into
a grid having a number for each row and column. Then any two values
are chosen from a random number table, which forms a coordinate pair
for defining the location of observation. The coordinate pairs are then
plotted on the image which shows the locations where sampling is to be
carried out.

• Stratified Random Sampling Pattern: This sampling pattern is same as

purely random sampling pattern except the fact that observation sites are
distributed (i.e. stratified) across all the categories (i.e. strata).
Stratification allows distribution of observation sites to categories in
proportion to their size significance as shown in Fig. 9.1b.

(a) Random Sampling (b) Stratified Random Sampling

(c) Systematic Sampling (d) Systematic Stratified

Unaligned Sampling

(e) Clustered Sampling

Fig. 9.1: Different types of sampling patterns employed in ground truth data collection
in remote sensing based studies
52
• Systematic Sampling Pattern: In this strategy, sample points are Ground Truth Data
Collection
distributed at equal intervals according to a specific strategy (Fig. 9.1c)
introducing systematic pattern. This pattern is useful in cases where it is
necessary to represent all the regions within the study area.
• Systematic Stratified Unaligned (non-aligned) Pattern: This sampling
pattern combines features of both stratified and systematic sampling
strategies while preserving an element of randomness. In this strategy, the
study area is divided into uniform cells by means of a grid (Fig. 9.1d).
• Clustered Sampling Pattern: In areas where accessibility is a concern or
You should collect ground
when there are other practical constraints such as budget limitations to
data at locations which are
employ other sampling methods, observations are clustered at nearby homogeneous and having
places. Such sampling pattern is known as clustered sampling pattern. In an areal extent equal to
this pattern, some sites are chosen within the study area and these become about 2x2 pixels of the
the centres to determine locations of additional observation sites located data being used to
understand pure spectral
nearby (Fig. 9.1e). These additional observation sites are called as
signatures and avoid any
satellite points. Locations of the centre points can be chosen randomly or mixing arising due to
in some specific areas. The tendency of similarity in measurements at one heterogeneity.
observation site to other sites is described by the term spatial
autocorrelation.
It is really difficult to say which method is best for ground truth data
collection. Some studies find that random sampling pattern performed best for
agriculture, stratified random sampling best for rangeland whereas systematic
and systematic stratified unaligned pattern performed well for forests. It can be
suggested that any one sampling strategy cannot be used for all types of
landscape. Each of these methods has its own advantages and disadvantages as
given in Table 9.1.

Table 9.1: Comparison of the five types of sampling patterns

Sampling Advantages Disadvantages
method
Random • Probably the most powerful • Sample points may not be uniformly
sampling strategy distributed in the entire study area and
• Minimum human bias may be clustered at one place
• Points may not be distributed
proportionately across all the
categories
• Categories having small areal extent
may be undersampled or may not be
represented at all
• Points may fall at or near the
boundaries of categories, which should
be avoided to use for accuracy
assessment purpose

Stratified • All the categories are

random represented as per their size
sampling significance
Systematic • Useful when it is necessary • Such a regular orientation of sample
sampling to ensure that all regions sites may introduce bias due to some
within the study area are linearity of patterns in the landscape
represented
53
Image Interpretation • Selection of the starting point
predetermines locations of all
subsequent observations
• Data derived using this strategy does
not meet assumptions of the
underlying statistics used in image
analysis

Systematic • Assures even distribution of • If grid size is small relative to ground

stratified sample points in the study features and large number of cells are
unaligned area present, element of randomness may
sampling be lost
• All categories are represented

Clustered • Useful in hard to access • There is no thumb rule regarding the

sampling terrains distance of sample site from its node
• Allows the analyst to make • There may be problem of spatial
most of accessible sites autocorrelation leading to false
• Data collection time may be accuracy of maps
reduced if lower number of
centre points (nodes) are
chosen, however, many
sample sites may be visited
at each nodal location

Spend Check Your Progress II

5 mins
1) Name the various types of sampling patterns.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................

9.3.4 How to Reach to the Ground Truth Sites and Locate

Yourself on the Ground?
The next step is to reach to the sampling sites for ground truthing after you
have determined where to collect data. While planning how to reach ground
truth sites, you should keep in mind the time difference between the date and
time of ground truthing campaign and when available maps were generated.
Then you should look for reliable locational information such as permanent
landmarks before making observations which will help you to identify exact
locations of your ground truth sites and also able to correctly find those
locations on remote sensing data. However, you may find that most of the
times it is difficult to exactly locate those sites on remote sensing data because
of either scale difference or unreliable locational information. GPS receiver
enables you to navigate with confidence to a particular location on the ground,
which is derived from a map or a georeferenced image.
54
9.3.5 How to Collect? Ground Truth Data
Collection
Once you have located the sampling sites, it is time to make observations. It is
critical to know how to collect data because usefulness and accuracy of your
observations depends on how you collect data. You may find scores of
literature regarding how to collect data but you would find that no single
method can be universally applied for all types of studies. Finding appropriate
method for your project objectives is crucial so that you can accordingly make
observations and take measurements.

You should select a method for measurements which (in your opinion) is not
only appropriate for that kind of study but also acceptable to subject matter
experts and professionals in the area. Whichever method you choose for
making measurements, you should ensure that
• method provides appropriate representative data
• data is in a form which could be understood and interpreted by other team
members
• method is defensible among the peers and
• data collected and final outcome should be meaningful to end user.
The process of data collection is generally carried out in the following two
broad steps:
• selecting appropriate approach and making measurements and
• noting observations and recording measurements.
Let us now discuss more about the two steps one by one.

Selecting Appropriate Approach and Making Measurements

After locating yourself on ground truth site, you should now make
observations selecting an appropriate method. You have studied in subsection
9.3.2 that type of data collected can be categorised into qualitative and
quantitative data. Acquiring qualitative data is comparatively easier as you just
need to observe and identify features on ground and relate it with the remote
sensing data. Collection of quantitative data is comparatively difficult and
more time consuming. For collecting quantitative data particularly for
ecological applications, quadrats are extensively used. Quadrats are plots of a
fixed size in which number and composition of features, their size and shape
are measured. The plots are usually square or circular in shape. In quadrat
sampling, size of the quadrats is a critical consideration. You should determine
size of the quadrats based on the size of the objects being measured.

The quadrats are designed for estimating density of objects, i.e. number of
objects per unit area and their composition. Using quadrats enables to carry
out more intensive measurements only over smaller spatial extents and
quadrats can also be made permanent so that you can revisit the location.
However, fixed quadrats approach has some limitations.

In some cases, different type of sampling strategy is followed such as line

transact, belt transact, etc. which are considered as more practical, efficient
and relatively inexpensive for many biological populations. In line transact
55
Image Interpretation approach, a line is marked on the ground along a tape or rope and observations
about different features crossing line are recorded at either even interval or at
irregular interval as in Figs. 9.2a and 9.2b. In belt transact method,
observations are recorded for a narrow strip of land as shown in Fig. 9.2c.

(a) (b) (c)

Fig. 9.2: Different types of transact employed for sampling and recording observation.
(a) Shows an evenly distributed line intercept transact, (b) an unevenly
distributed line intercept transact and (c) a belt transact

From Fig. 9.3, you can make out that it is comparatively easy to make
measurements on land than coastal areas or underwater. On land, one has to
locate and reach predetermined positions on ground and make measurements
whereas in the coastal area one has to reach predetermined locations based on
tidal conditions. For making underwater measurements one has to reach the
sampling sites by a boat or ship and then make measurements either from
water surface or at the sea floor.

(a) (b) (c)

(d)
Fig. 9.3: Noting observation and recording measurements (a) on land, (b) in shallow
intertidal region, (c) in coastal waters and (d) a typical underwater survey
wherein an observer moves along a transact and records information

56
Noting Observations and Recording Measurements Ground Truth Data
Collection
Observations made in the field must be recorded in a manner that others can
interpret them without difficulty especially, if the image analyst is not the
same person who took observations. It is important to note that all the
information recorded is to the point and in clear language and nothing should be
left for data analyst’s imagination. Apart from information about the features of
interest, environmental parameters that could affect features of interest should
also be recorded. Another important point is to measure the spectral response of
features under different environmental conditions. Further, you should also
ensure that data recorded are suitable for future references too.
Recording of the above mentioned information at multiple locations requires a
systematic approach. And for the purpose, it is useful to prepare a form for the
purpose. These forms are known as ground truth data collection sheet or field
forms. Ground truth data sheets or field forms may vary for projects having
different objectives. However, this is a great aid to field personals to maintain
uniformity and consistency for multiple observations. When you design and
prepare a ground truth data sheet for your project, you should remember that it
covers all the items which are required to be collected on ground to meet
project objectives.
You will get an idea on how to prepare a field form (ground truth data sheet)
for your own project objectives from the three samples of field forms given in
Figs. 9.4 to 9.6.
Sample #1
Proforma of Ground Truth Data Collection Sheet
(For Mapping Land use/Land cover Features)
General Information
Sample No.: Observer’s Name:
Date: GPS Reading: N/S
Time: E/W
Location: Survey of India Topographical Map No.:
Vill.: Dist.: State: Landmark:
Accessiblity: Size of Ground Truth Site: m2

Ground Photographs: Topography:

N Photograph No. Direction
Elevation m
W E Slope: %
Aspcet:
S

Observations-Land use/Land cover:

Major Unit Sub Unit Image Feature Remarks

Any Other Relevant Observation Including Field Sketch:

Fig. 9.4: Proforma for ground truth data collection for mapping land use/land cover features
57
Image Interpretation Sample #2
Proforma of Ground Truth Data Collection Sheet
(For Biodiversity Related Studies)
General Information
Sample No.: Observer’s Name:
Date: GPS Reading: N/S
Time: E/W
Location: Landmark:
Vill.: Dist.: State:
Accessibility: Survey of India Topographical Map No.:
Topography: Size of Ground Truth Site:
m2
Slope: Image/map feature:
Aspect: Image Characteristics:

General Observations: Ground Photographs:

Wild life (including evidence): N Photograph No. Direction
Evidence of Disturbance, if any:
W E
No. of Trees:
Quadrat size: x m S

Observations:
Sr. Species DBH Crown Height Phenologic Importance Remarks
No. width status
1
2
3
4
5
6
7
8
9
10
DBH-Diameter of trunk at breast height,
Any Other Relevant Observation Including Field Sketch:

Fig. 9.5: Proforma for ground truth data collection for biodiversity or forest related studies

What to Do if You Do Not Have Sufficient Number of Observations?

What should you do if your observations are either incomplete or you are not
able to go to the field. In such case, you may need to take help of other
resources available. For this purpose, available thematic maps, topographic
maps, aerial photographs and even online resources can be used for collecting
information about the features of interest, though, these maps can never
replace the need for ground truthing. However, in the absence of sufficient
ground truthing data, internet resources such as ISRO’s Bhuvan or Google
Earth could prove as a good source of information for validating your outputs
particularly when your project involves use of medium scale remote sensing
data. There are some other websites which can be used to extract information
that can be used as a field validation data. Some of them are given below:
• http://bhuvan.nrsc.gov.in/bhuvan
• www.wikimapia.org
• http://maps.google.com
58
• www.mapslive.com Ground Truth Data
Collection
• http://glovis.usgs.gov
• www.itouchmap.com
• www.terraserver.com/view.asp?tid2.

Sample #3
Proforma of Ground Truth Data Collection Sheet
(For Field Radiometer Related Measurements)

General Information
Sample No.: Observer’s Name:
Date: GPS Reading: N/S
Time: E/W
Location: Survey of India Topographical Map No.:
Vill.: Dist.: State: Landmark:
Accessiblity: Size of Ground Truth Site: m2

Ground Photographs: Geometry of observer’s position relative to

N
Photograph No. Direction incident radiation:

W E

Instrument Observations
Height of instruments above ground: Height of instruments above targe:
Sr. Scan No. Target Time Sky Nearby features that Remarks
No. Condition may affect incident
radiation

DBH-Diameter of trunk at breast height

Any Other Relevant Observation Including Field Sketch:

Fig. 9.6: Proforma for ground truth data collection for recording field radiometer
related observations
Check Your Progress III Spend
5 mins
1) List the broad steps of data collection process.
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
......................................................................................................................
...................................................................................................................... 59
Image Interpretation 9.3.6 What to Carry While Going for Data Collection?

It is difficult to give here an exhaustive list of all things to carry while going to
ground because they would vary depending upon project objectives. However,
an indicative list of requirement is given below:
1) Baseline maps
2) Georeferenced images
3) Interpreted outputs/maps
4) Map cases
5) Letter of permission/approval to visit the study area
6) Stickers and flags for labelling samples
7) Ground truth data sheets and forms
8) Field spectrometer with required accessories
9) GPS and camera with extra batteries/charger
10) Other instruments for measuring distance, elevation, as required
11) Measurement tape, compass and scale
12) Sample bags and/or bottles
13) Field notebooks, pens, pencils, markers, sharpeners and erasers

Further, you may need to carry hand lenses, secchi disk, soil testing kit,
Munsell colour charts, camping items, underwater housing/casings items, etc.
depending upon project objectives.

9.3.7 Estimating Cost of Data Collection

Collection of ground truth data is an expensive task. Costs associated with all
components related to ground truthing is vital and you should estimate them in
advance to ensure that appropriate budget is available. Costs associated with
ground truthing components can be categorised into fixed cost and variable
cost.

You can define fixed costs as costs which are independent of the types of
images used, duration of survey, sampling strategy, number and distribution of
sampling sites. It includes costs of the following:
• generation of base map or purchase
• field equipments and instruments
• computing equipments and storage devices
• image processing software and
• printing device and related consumables.

You can define variable costs as costs which would vary with the types of
images used, duration of ground truthing, sampling strategy, number and
distribution of sampling sites. It includes costs of the following:
60
• hiring cost of manpower and transport Ground Truth Data
Collection
• areal extent and accessibility of the study area
• sampling method employed and
• level of data required.

9.3.8 Do’s and Don’ts

Field work is not an easy job and you may face problems during data
collection. In order to avoid such problems you must consider certain do’s and
don’ts along with safety concerns listed below:
• do not alter the notes already written on your field notebooks once you
are out of the field. Carry with you all the approvals to visit the study area
• if permission for access to certain locations cannot be obtained, other
locations analogous to those locations need to be considered
• behave politely with local people you meet during your field work
sessions. You may not be allowed to enter into a specific area if these
people find your behaviour suspicious
• carry your identity proof with you and first aid box
• make note of weather predictions so that you would be prepared for what
to expect
• never work alone particularly in an unknown terrain. Always work in
groups
• wear appropriate clothing as per the weather conditions. Clothes should
be preferably of yellow, orange or red colours. These colours could be
easily seen from a distance so that you could be traced and located if you
get into any trouble
• for field work in desert, costal and in high rainfall areas you need to be
prepared accordingly such as with extra water, glucose tablets, snacks,
antibiotics, anti venoms, and precaution for mosquitoes, humidity and so
on and
• follow the instructions of your instructor or team leader.

The points mentioned above give you an idea about the problems one may
face during ground truthing.

9.4 WHAT TO DO ONCE THE DATA IS

COLLECTED?
After ground truth data has been collected you plot the locations of sampling
sites recorded by GPS on remote sensing images of the study area using
software. You should note that there could be shifts in the locations of ground
truth sites and the image either due to inaccuracies in the GPS derived
coordinates or due to error in georeferencing of the image being used. The
shift may vary from one GPS model to others depending upon its accuracy.
After plotting the locations, the simplest task is to use ground truth
61
Image Interpretation information for those pixels to label them in classes. If ground data consists of
spectral data then you can make a comparison of spectral signatures collected
from ground and signatures derived from remote sensing data. The spectral
signature can be used as an input for developing models.

Another important task is accuracy assessment using ground truth data.

Though you would be studying about accuracy assessment in Unit 14 of
MGY-002, it would be appropriate to know here how ground truth data is used
for accuracy assessment. If ground truthing is carried out only once during the
entire duration of the project then ground truth data collected is divided into
two parts, one part is used as an aid for image interpretation and second part is
used to assess accuracy of maps generated. However, if ground truthing is
carried out at least twice during the entire duration of the project then the
entire ground truth data collected for the first time is used as an aid for image
interpretation and then another ground truthing is carried out to assess the
accuracy of maps generated.

9.5 ACTIVITY
You have read about ground truthing, types of data and also how to collect
ground truth data.

First, you can visit a website offering freely downloadable remote sensing
images of your area. You can also use images from Google Earth, wikimapia
or other such websites. Now choose an area of image, which you want to take
up for ground truthing and then take a print out of that small part of the image.
Now you can try to identify key features from the image and trace the
boundaries of those features on a tracing sheet. Label names of all features you
have identified and traced (mapped) and also put coordinates of map. After
interpreting, you visit area the image belongs to and also carry with you the
image, traced sheet (map), a notebook and a pen/pencil. First of all, you need
to locate yourself on ground with the help of coordinates marked in the map
and then compare features you have mapped and relate it with feature as seen
on the ground. Write information such as ground truth site number, locational
information, what you observe on the ground, and what that feature mapped
has been labelled as in your notebook. Make such observations at 5-6
locations and note all relevant information. After completing task, compare
map and related image features with information collected on ground. You can
modify labels and boundaries of features on maps based on ground truth data.

9.6 SUMMARY
Let us summarise what you have studied in this unit:
• Ground truth data refers to the data collected on surface of the Earth. It is
complementary to the remote sensing data. It helps to link image data to
ground reality.
• Ground truth data includes (i) attributes or measurements describing
ground conditions (ii) observations must be linked to locations and (iii)
observations must also be described to time and date.
• Ground truth data can be grouped as nominal, biophysical, field
62
 radiometric and locational information data. Ground Truth Data
Collection
• Ground truth data collection should be planned based on project
objectives, nature of features being studied and their geographic
distribution and available budget.
• Timing of ground truth data collection depends upon nature of features
being studied, environmental conditions and the stage of interpretation.
• Choosing a sampling pattern depends upon objectives and accessibility of
the area, nature and characteristics of the features being studied.

9.7 UNIT END QUESTIONS

1) What is ground truth data? List out its essential elements.
2) What is sampling pattern? How would you decide to choose a particular
sampling pattern?
3) How would you locate yourself on the ground?
4) What are the things one should carry while going for data collection?
5) How would you estimate cost of ground truth data collection?

9.8 REFERENCE
• Campbell, J.B. (1996), Introduction to Remote Sensing, The Guilford
Press, New York.

9.9 FURTHER/SUGGESTED READING

• Campbell, J.B. (1996), Introduction to Remote Sensing, The Guilford
Press, New York.
• McCoy, R.M. (2005), Field Methods in Remote Sensing, The Guilford
Press, New York.

9.10 ANSWERS
Check Your Progress I
1) There are two main purposes of ground truth data collection, firstly, it is
used as guide for image interpretation and secondly, it is used to validate
information derived from image data.

Check Your Progress II

1) Types of sampling patterns are simple random, stratified random,
systematic, systematic stratified unaligned and clustered sampling
patterns.

Check Your Progress III

1) Broad steps of data collection process are (i) selecting appropriate
approach and making measurements and (ii) noting observations and
recording measurements.
63
Image Interpretation Unit End Questions
1) Refer to section 9.2
2) Refer to subsection 9.3.3
3) Refer to subsection 9.3.4
4) Refer to subsection 9.3.6
5) Refer to subsection 9.3.7

64