00:00:00
all right today we're going to become
00:00:01
familiar with this spatial analysis
00:00:03
which is one of the fanciest topics in
00:00:05
GIS first we will learn what is the
00:00:08
spatial analysis and then some basic
00:00:10
operations and spatial analysis
00:00:13
including selection and classification
00:00:17
so far we have talked about the
00:00:21
definition of GIS GIS data models vector
00:00:25
data model and raster data model how to
00:00:28
represent your graphic features and also
00:00:32
locations with coordinate geographic
00:00:35
coordinate system and projected
00:00:37
coordinate system how to manage
00:00:39
attributes into database and how to
00:00:43
produce GIS data from scratch
00:00:46
collecting GPS waypoints as a control
00:00:49
points then geofencing scan map and
00:00:52
finding and digitizing features so all
00:00:56
of them all of these topics that we
00:00:59
covered for about the data collection
00:01:02
and data preparation data production
00:01:05
from scratch and data representation as
00:01:09
map now we are ready to move to the next
00:01:11
step which is spatial analysis we have
00:01:15
the GIS data like census data land use
00:01:21
data land cover Road meant for
00:01:25
typography data such as digital
00:01:27
elevation model disease data such as
00:01:31
disease count prevalence incidence and
00:01:33
the next step is to combine these data
00:01:36
and turn the data into knowledge to
00:01:39
solve real world problem for example
00:01:42
where is the best place to build a new
00:01:45
hospital based on several criteria
00:01:47
factors like slow proximity to high
00:01:50
population proximity to road should be
00:01:53
combined or where are the high risk
00:01:55
areas for disease outbreak we can
00:01:58
allocate resources budget to these areas
00:02:00
to control the outbreak which route is
00:02:03
the shortest route between your house to
00:02:05
the University so these are knowledge
00:02:07
that we can extract from the data so
00:02:11
what is the spatial analysis technique
00:02:13
the spatial analysis is the application
00:02:16
of one or more GIS operation in order to
00:02:19
solve real-world problem such as what is
00:02:22
the shortest path between your house and
00:02:24
department by car how many people are
00:02:27
living within one kilometer around the
00:02:29
hazardous situ risk assessment or or
00:02:33
health disparity which areas have lowest
00:02:36
access to healthcare or another
00:02:38
real-world application in public health
00:02:40
according to Virginia Department of
00:02:42
Health is Virginia coupled GIS mapping
00:02:46
many spatial analysis to identify areas
00:02:49
where infant mortality rates are the
00:02:52
highest the extent of racial and ethnic
00:02:55
disparities in infant deaths the
00:02:57
underlying causes of those infant deaths
00:02:59
and how to best intervene so these are
00:03:02
knowledge you can extract from the
00:03:04
prepared data to support your
00:03:06
decision-making to make decision based
00:03:08
on the fact technically a spatial
00:03:13
analysis is a chain of GIS operations a
00:03:16
chain basically means the output from
00:03:18
one joyous operation is served as the
00:03:20
input of the second GIS operation and so
00:03:23
on so operations are linked together to
00:03:26
solve the real-world problems so here
00:03:30
for example the input layer are we have
00:03:33
an input layer a special analysis is
00:03:36
done on the input layer so we will have
00:03:38
an output layer the output layer will be
00:03:40
the input for the second special
00:03:42
operation and gives us the output layer
00:03:44
and this output layer serves as input
00:03:47
for the special operation and gives us a
00:03:50
final output layer so you can see the
00:03:52
operations are linked together to solve
00:03:54
the real-world problem and by chaining
00:03:59
GIS operation you will have a
00:04:01
complicated GIS flowchart like this and
00:04:04
a couple of input layers with many
00:04:07
operations linked together it is called
00:04:10
a GIS model in the future lecture we
00:04:13
will build sophisticated GIS model in
00:04:16
ArcGIS and solve complicated real world
00:04:19
problems that are not easy to solve
00:04:22
without GIS
00:04:24
so basically there are three categories
00:04:27
of Jonah's operation we will talk about
00:04:30
a little today we just scratch the
00:04:33
surface and with more details in the
00:04:36
future lectures so GIS operations can be
00:04:39
divided into three categories local
00:04:41
operations neighborhood operations and
00:04:43
global operations local operations
00:04:46
basically mean the output is determined
00:04:50
based on the input in the same location
00:04:53
so for input and output are for one
00:04:56
location for instance this gray polygon
00:05:02
is the state of Utah we want to
00:05:04
calculate population density of this
00:05:07
state and the output here only depends
00:05:11
on the population size and also the area
00:05:14
of that state so we only need
00:05:16
information about Utah State to
00:05:18
calculate population density of the same
00:05:21
estate so that's called local operation
00:05:24
the second type is called neighborhood
00:05:26
operation and this operation uses data
00:05:29
from both input location plus the nearby
00:05:32
or neighborhood locations to determine
00:05:36
the output value so that's why it's
00:05:38
called neighborhood operation for
00:05:40
example let's say let's say you want to
00:05:45
find out the number of adjacent say so
00:05:48
the Utah State okay or in other words
00:05:51
how many estates share boundary video da
00:05:54
in this case we not only consider the
00:05:58
location of Utah State but also the
00:06:00
surrounding estates as well to calculate
00:06:02
how many edges in the state's to Utah
00:06:04
State which is one two three four five
00:06:08
six and out would will be six right so
00:06:13
that is called neighborhood operation
00:06:15
and neighborhood operation involves
00:06:17
neighbors of the city region from
00:06:20
calculation for example for California
00:06:22
its shares border with one two three so
00:06:26
the output will be three and depends on
00:06:29
the neighbors third type is global
00:06:33
operation which means this operation
00:06:35
uses the
00:06:36
value from the entire input layer to
00:06:39
determine each output value for example
00:06:42
here here is let's say we want to rank
00:06:47
Western estates by their total
00:06:51
population for the year 1990 in this
00:06:54
case we not only consider the population
00:06:56
size of for example Utah but also the
00:07:00
population size for all of the Western
00:07:04
estates all of this study area and then
00:07:06
we can rank them in order and find out
00:07:09
Utah and Chile has a rank order of six
00:07:12
okay and California has the largest
00:07:15
population ranked number one among the
00:07:18
other state the second one is Washington
00:07:21
and so on so global operation involve
00:07:25
entire study area for calculation GIS
00:07:35
operations our building blocks for
00:07:38
complicated spatial analysis or spatial
00:07:41
analysis consists of several joist
00:07:43
operation which aims to address
00:07:45
real-world issue basically there are
00:07:47
four types of GIS operations selection
00:07:50
classification and proximity and overlay
00:07:53
which are applicable for both vector and
00:07:56
raster data set we're going to talk
00:07:58
about the operations one by one and in
00:08:01
today's class we will talk only about
00:08:04
selection and classification and
00:08:07
proximity and overlays for the future
00:08:10
lectures let's start with the selection
00:08:15
so selection is a GIS operation that
00:08:19
creates a subset of data that satisfies
00:08:22
certain criteria so there are two types
00:08:25
of selection selection my attribute and
00:08:27
selection my location selection my I
00:08:30
distribute that we already have worked
00:08:32
with we use that to select a specific
00:08:34
type of cancer based on location and it
00:08:38
selected specific records or feature
00:08:40
based on attribute values but the
00:08:44
selection by location selects record
00:08:46
based on a special relationship like
00:08:49
in composition interception at Jason's
00:08:52
adjacent or or in general of topological
00:08:56
relationship that we talked about before
00:08:59
so selection by location is only
00:09:02
applicable for vector data model because
00:09:05
raster they talk and do not have any
00:09:08
spatial topological information so let's
00:09:13
start with the selection by attribute
00:09:15
for vector data there are a number of
00:09:18
operators you can use like and not or
00:09:21
greater than less than equal to and so
00:09:24
on to write your statement or query so
00:09:28
we can use these operators to build a
00:09:30
query a statement was said to specify
00:09:33
the selection criteria so we've won with
00:09:36
that in the previous labs so basically
00:09:39
in select by attribute GS applies the
00:09:42
query on each feature and then compare
00:09:44
that reviews of each feature with the
00:09:47
criteria the feature or records that
00:09:49
meet the criteria will be selected for
00:09:52
example here we want to select all
00:09:54
counties with attribute name of the
00:09:57
state of Vermont so it is based on the
00:10:00
attribute table or we want to select all
00:10:03
counties that their attribute name is
00:10:05
not New York so this operator is shows
00:10:10
not and/or we want to select all
00:10:13
counties with areas larger than or equal
00:10:16
to 1,000 square miles so you can see
00:10:20
that so it's based on the area based
00:10:22
based on the attribute table there is a
00:10:24
field that the name of area so based on
00:10:26
that one we can select all counties or
00:10:29
select the counties that population
00:10:31
density is less than 250 percent per
00:10:34
square mile
00:10:35
so these are the selected counties
00:10:37
selection by attribute can also be
00:10:40
applied to raster data so let's say we
00:10:43
have a very simple raster dataset which
00:10:46
describes land you stop for a study area
00:10:49
one represents agriculture two
00:10:51
represents forests and three represents
00:10:54
urbanized area let's say we want we only
00:10:57
want to tease out the urbanized areas
00:11:00
okay we can define the SQL statement
00:11:03
like land use type across the tree and
00:11:06
then GIS will read this criteria and
00:11:09
then compare that compared every single
00:11:12
pixel one by one so let's start with the
00:11:15
first cell the first cell is one and one
00:11:19
doesn't satisfy the criteria because
00:11:22
language type is not equal to three so
00:11:24
the output will be 0 which means false
00:11:28
and then we go to the next cell the next
00:11:32
cell again is 1 and doesn't satisfy so 0
00:11:35
the second the third one is 2 and
00:11:38
doesn't satisfy it's not 3 so it's 0 so
00:11:42
until we reach this cell this cell is 3
00:11:47
& 3
00:11:48
land use type is Erinn area so it
00:11:50
satisfies the criteria so the output
00:11:53
will be 1 so 1 means that criteria is
00:11:57
satisfied and so that's how GIS that's
00:12:01
how can be lower scan is scan and do
00:12:04
this like by attribute or the raster
00:12:06
dataset okay so keep in mind that the
00:12:09
output of selection by attribute for
00:12:12
raster dataset is always a binary great
00:12:14
it's oh it's either 0 or 1
00:12:18
okay or how about this one line is type
00:12:21
greater than 1 so it compares each cell
00:12:24
to the criteria and if satisfies this
00:12:27
criteria the result would will be 1
00:12:29
otherwise it will be 0 so again the
00:12:32
output is only binary raster so let's
00:12:38
move on to the second type of selection
00:12:40
which is selection by location so here
00:12:43
our selection criteria is not defined
00:12:45
based on attribute information it's
00:12:48
defined based on a spatial relationship
00:12:50
or topological information special
00:12:53
relationship between two features we
00:12:56
have talked about before include the
00:12:57
intersection of two lines or adjacency
00:13:01
between two polygons or composition of
00:13:03
polygons or line or containments so
00:13:06
these are spatial relationship and
00:13:09
selection by location usually involves
00:13:12
two or
00:13:14
more layers as the input so we need at
00:13:17
least two layers for the selection by
00:13:19
location for the vector data let's see
00:13:23
one example of selection in my location
00:13:25
based on adjacent selection so we want
00:13:29
to know which assets are adjacent to
00:13:31
Missouri and so we have a shapefile of
00:13:35
Missouri and a shape far away all of the
00:13:37
US states now we're not looking at
00:13:40
attribute of the Missouri assay but we
00:13:43
look at the spatial relationship between
00:13:46
Missouri estate and the surrounding
00:13:48
states and then the surrounding polygon
00:13:50
can be teased out like this so our
00:13:54
containment selection is another
00:13:56
topology our relationship for instance
00:13:59
which estates contain the Mississippi
00:14:02
River so this is the Mississippi River
00:14:06
okay and and we have it as a shapefile
00:14:12
and also we have used US states as
00:14:15
another shapefile and all states that
00:14:17
contain part of Mississippi River the
00:14:21
entire state will be selected out so
00:14:23
these gray estates okay all of these
00:14:26
gray says they contain part of
00:14:29
Mississippi River
00:14:30
so in selection by location we're
00:14:32
looking at the spatial relationship
00:14:34
between the features for example
00:14:36
adjacency or containment sometimes its
00:14:41
intersection sometimes it's composition
00:14:43
so this is called selection by location
00:14:45
we're not looking at the attribute table
00:14:47
but we're looking at the relationship
00:14:50
between the features and
00:14:53
ArcGIS provides a user-friendly
00:14:55
interface to help us to construct SQL
00:14:59
statement to a specify selection
00:15:01
criteria and this dialog helps us to
00:15:05
define how to select features based on
00:15:08
the attribute or based on the spatial
00:15:13
relationship so here is like by
00:15:15
attributes and here is like by location
00:15:18
we already have work birds like by
00:15:21
attribute and in this lab we will also
00:15:23
work with select by location so in
00:15:26
summary for select by
00:15:28
beautiful Victoria Talley apply the
00:15:29
criteria on each feature if the features
00:15:33
satisfy the criteria it will be selected
00:15:35
and select by attribute for us there we
00:15:38
apply the criteria on each cell each
00:15:40
pixel and if the feature satisfies the
00:15:43
criteria it will be 1 otherwise it will
00:15:46
be 0 and select by location has nothing
00:15:52
to do with the attributes it's only
00:15:54
applicable for vector data set it cannot
00:15:57
be used for the raster dataset because
00:16:00
it's based on the relationship or is
00:16:02
based on the topological relationship
00:16:04
and topological relationship can only be
00:16:06
defined for the vector data model not
00:16:10
for the raster data model another widely
00:16:13
USG operation is classification
00:16:16
classification is also known as
00:16:18
reclassification or recoding and this
00:16:21
operation will summarize the features
00:16:23
into several groups based on predefined
00:16:26
conditions so let's say we have an input
00:16:29
raster layer and we want to simply use
00:16:34
new values to replace all values based
00:16:37
on a predefined condition and here is a
00:16:39
lookup table and lookup table basically
00:16:43
defines how old values will be replaced
00:16:46
by new values all values between 1 to 3
00:16:50
will be replaced by 5 and all values
00:16:54
between 3 to 7 will be replaced by new
00:16:57
values 3 and so on so for example first
00:17:02
cell is has a value of 3 which is old
00:17:07
value then you look up look at the
00:17:09
lookup table the cells falls into this
00:17:12
category right so the new value should
00:17:15
be 5 okay in the output and output will
00:17:18
be saved as 5 here for this cell and if
00:17:22
you look at so the second one is also
00:17:25
the same 5 the third one is 19 and 19
00:17:28
falls into this category and the new
00:17:31
value for this value for this for 19
00:17:34
will be 5 again so we can continue this
00:17:38
process for every single cell 1 by 1 2
00:17:42
reclassify this image okay so this is
00:17:45
called classification or
00:17:47
reclassification process which is widely
00:17:49
was for summarizing and displaying
00:17:52
datasets so there are many examples or
00:17:56
applications for classification for
00:17:59
example we can classify zip codes into
00:18:02
low crime rate high crime rates so two
00:18:06
categories we can summarize cancer type
00:18:09
such as brain cancer into medium and
00:18:14
high and low rate of the cancer or we
00:18:19
can summarize habitat suitability for
00:18:23
red fox to very high high medium low
00:18:26
very low suitability so you can see
00:18:29
there are numerous applications for
00:18:31
classification so now let's talk about
00:18:35
classification for all attribute size we
00:18:39
have talked about different types of
00:18:40
attributes before which were nominal
00:18:43
ordinal interval and ratio so we are
00:18:46
start with nominal and ordinal
00:18:48
attributes nominal and ordinal
00:18:51
attributes are also referred to as
00:18:54
categorical variable or categorical
00:18:57
attributes for classification of
00:19:00
categorical tribute is very easy we
00:19:03
should use a lookup table for
00:19:04
classification like this and lookup
00:19:07
table basically defines how old values
00:19:10
will be replaced by new values for
00:19:12
example let's say we're going to recode
00:19:14
all the states in the west of main
00:19:18
branch of Mississippi River as one and
00:19:21
an all estates in the east of river as
00:19:25
zero so here is a classification table
00:19:28
all values are categorical values right
00:19:33
so these are just names and new values
00:19:37
are specified as either one or zero so
00:19:41
here is their classification results
00:19:43
okay so that's what categorical
00:19:46
attributes summarize or group the data
00:19:48
now let's move on to the interval and
00:19:53
ratio attribute we
00:19:55
are also referred to as numeric
00:19:57
attributes which is more sophisticated
00:20:02
so for example let's say we have a study
00:20:06
area this is this is our study area this
00:20:08
is a community with more than 1,000
00:20:12
polygons and each polygon has a
00:20:15
population and population size ranges
00:20:18
from 0 to 500 5,000 and 100 and let's
00:20:25
say you want to reclassify our study
00:20:28
area into three categories based on
00:20:31
population size - low medium and high
00:20:34
population size so three categories mean
00:20:37
we need to define or determine two
00:20:40
cutoff values okay a and B so below a is
00:20:45
assigned to low category between a and B
00:20:51
polygons will be assigned to their own
00:20:53
medium population size and above B we
00:20:57
consider that consider this polygon as
00:20:59
high population size so question is how
00:21:03
to determine the cutoff values a and B
00:21:05
and the answer is we need some automatic
00:21:09
classification method to help us to
00:21:11
determine or define threshold values
00:21:14
there are a variety of ways for
00:21:17
classification and the first method is
00:21:19
called equal interval classification
00:21:22
which means the range of each category
00:21:23
is going to be the same or equally
00:21:26
Islita range from 0 to 5000 and 133
00:21:30
which is the maximum population so you
00:21:34
can see the entire study area is divided
00:21:36
into three categories and from low with
00:21:40
white color medium with gray color and
00:21:44
high with black color and if you look at
00:21:46
the range of each category from minimum
00:21:49
to maximum ranges of each class or
00:21:51
category is equal that's called equal
00:21:54
interval classification maybe this
00:21:58
picture is not really clear so let's see
00:22:01
the next picture the next slide that
00:22:04
percentage of population under
00:22:07
five so has been classified into five
00:22:11
categories and when you look at the
00:22:14
ranges the ranges are between 3 to 6 6
00:22:17
to 9 9 to 12 12 to 15 15 to 18 so the
00:22:22
ranges are equal or they have equal
00:22:26
interval okay so that's the first type
00:22:30
of classification the second method that
00:22:34
help us to determine the cutoff value is
00:22:37
called quantile classification so in
00:22:40
this classification method the threshold
00:22:42
or color values are set so that each
00:22:45
category is going to have the same
00:22:47
number of special features special
00:22:50
features the same number of features for
00:22:53
example in this color and the same
00:22:55
number of features in this color and so
00:22:58
for so on so this classification method
00:23:01
is suitable for linearly distributed
00:23:03
data so you can see the categories will
00:23:06
have however their range for each
00:23:08
categories are different within each
00:23:12
category the number of features is
00:23:14
almost the same or the polygons are
00:23:17
approximately the same that's called
00:23:19
equal number classification or quantile
00:23:22
classification so let's see an
00:23:25
application of contour classification in
00:23:27
GIS for example we want to rank in US
00:23:31
states based on their areas into five
00:23:33
categories so we use quantile
00:23:36
classification if you generate five
00:23:39
classes this means that ten states will
00:23:42
reside in each class each class the
00:23:46
first class will have the ten largest
00:23:49
states in terms of land mass and the
00:23:52
last class or last category will have
00:23:55
the ten as small as the States in terms
00:23:57
of land mass so when you use the
00:24:00
quantile map classification with five
00:24:02
classes it will look like this map so it
00:24:06
is easy to see see that it is easy to
00:24:10
see that it states like Texas or
00:24:12
California or New Mexico are in the top
00:24:17
ten for
00:24:18
his closest aides are obviously their
00:24:22
smallest so quantile classification is
00:24:26
ideal when there is a order in the data
00:24:28
it is suitable for ordinal data and
00:24:32
third method method that helps us to
00:24:35
determine the cutoff value is called
00:24:37
natural break classification and this
00:24:40
might be the most widely used
00:24:41
classification technique this
00:24:43
classification is based on the histogram
00:24:45
and looks for the obvious or largest
00:24:48
gaps in the data so let's say in the
00:24:51
same study area this is a histogram of
00:24:55
the same study area and x-axis indicates
00:25:00
population size from zero to maximum
00:25:03
which was 5133
00:25:05
and the y-axis tells us the frequency of
00:25:10
the occurrence of the population which
00:25:12
means number of polygons you have for
00:25:15
this population size then if you look at
00:25:18
histogram here we can identify the
00:25:20
number of natural gaps for example here
00:25:22
is a gap and also here is another
00:25:25
largest gap in the data so the natural
00:25:29
break classification uses mathematical
00:25:32
formula to help us to find the largest
00:25:34
gaps in histogram and then we use these
00:25:37
gaps as cutoff values so technically it
00:25:40
minimizes variance within each class and
00:25:44
maximizes variance between the classes
00:25:46
the last classification method is called
00:25:49
manual or defined interval
00:25:51
classification method so the choice of
00:25:54
cutoff values is up to you and you
00:25:58
define intervals or cutoff values so we
00:26:05
talked about four different type of
00:26:07
classification method for numeric
00:26:09
attributes manual defined equal interval
00:26:14
quantile and natural break so you can
00:26:17
see for the same data you got different
00:26:19
visual results or different pattern
00:26:22
that's why people say maps can be
00:26:25
misleading if I change cutoff value for
00:26:28
each class I will get different result
00:26:30
if I change the class
00:26:31
vacation technique I will get different
00:26:33
results so the question is which one
00:26:36
gives us the best display of data and
00:26:39
here are some criteria or guidelines for
00:26:43
you to select the best classification
00:26:45
method so you always have to look at the
00:26:47
data when when do we use natural break
00:26:51
so we use natural breaks when the
00:26:53
attributes like I don't know population
00:26:56
size are distributed unevenly across
00:26:59
overall range of the data so then when
00:27:03
you look at the histogram of the data
00:27:04
you will see some pics the distribution
00:27:07
will be like this and in that case
00:27:11
natural breaks might be the best choice
00:27:14
so you choose numbers that best reflects
00:27:18
natural gaps within your data and how
00:27:22
about equal interval equal interval
00:27:24
classification is suitable when you want
00:27:28
to have all classes with the same range
00:27:31
for example if you want to display
00:27:32
category score every 1000 increments and
00:27:35
then interval is 1000 for every category
00:27:38
in that case you call interval is
00:27:41
suitable the quantile classification
00:27:46
basically produces the same number of
00:27:48
the geographic feature for every
00:27:50
category and the best time you can use
00:27:52
this classification is menu when your
00:27:55
attributes are linearly distributed
00:27:57
across the range which means that if you
00:27:59
draw out a histogram there is no pick or
00:28:04
basically it's flat for histogram or
00:28:06
probably there is a very it has a very
00:28:08
gentle slope and the manual
00:28:11
classification and the manual defined
00:28:13
classification can be used when you want
00:28:15
your classes to break at a specific
00:28:17
values so that's a guideline for
00:28:20
classification method so let's open up
00:28:24
our jaws to show you classification
00:28:27
techniques first of all we need to
00:28:30
download US sa foundries from tiger line
00:28:34
so if I here type tiger line
00:28:39
and then go to the tiger line website
00:28:44
the most recent data web interface and
00:28:54
states and equivalent submit download
00:29:00
national file and then let's extract it
00:29:21
and loaded into ArcGIS
00:29:39
we're going to classify this shapefile
00:29:44
based on the land area so if you open up
00:29:48
attribute table of their states there is
00:29:52
a field with name of a land I think this
00:29:58
area land and I'm going to classify it
00:30:02
based on the symbology symbology and its
00:30:10
quantity right because it's a numeric
00:30:13
value and then graduate color and base
00:30:19
under a land so here the classification
00:30:25
technique is natural breaks right and
00:30:27
here we have five number of the classes
00:30:30
if I click apply and then okay so I can
00:30:34
see that so these areas you have higher
00:30:36
so remember this picture okay so if I
00:30:40
here instead of natural break with five
00:30:42
classes I use I don't know click on
00:30:46
classify and then I select equal
00:30:49
interval okay and then okay apply okay
00:30:55
so you can see that chain the shape
00:30:58
completely changed right so depend on
00:31:01
the classification technique we have
00:31:03
different cutoff values and then we have
00:31:07
different Maps right so or we can change
00:31:11
number of the classes instead of five
00:31:13
classes we can change it to three
00:31:16
classes and then we get different
00:31:17
results right or here we have another
00:31:21
type of classification you can manually
00:31:23
define the break values here okay you
00:31:26
can change them or you can use equal
00:31:30
interval define interval natural break
00:31:33
or other type of classification
00:31:35
techniques and each one gives us
00:31:38
different maps so so you can create
00:31:42
numerous maps with the same data that's
00:31:44
why in elections or advertisement they
00:31:48
use maps to deceive people the data are
00:31:52
the same
00:31:52
and true but classification techniques
00:31:55
are different
