Beruflich Dokumente
Kultur Dokumente
Prof. S.K.Ghosh
Dept of Civil Engg
Analysis in GIS
Before starting any analysis, one needs to assess
the problem and establish an objective.
It is important to think through the process before
making any judgments about the data or reaching any
decisions;
Ask questions about the data and model; and
Generate a step-by-step procedure to monitor the
development and outline the overall objective.
Feature:
Data layer: A data layer is a data set for the area of interest in
a GIS, normally containing
data of only one entity type (i.e.,
points, lines, or areas).
Image: A data layer in a raster GIS is called as image.
Cell: An individual pixel in a raster is a cell.
Function: Function or operation is a data analysis procedure
or operation performed by a GIS.
Algorithm: A sequence of actions to be performed by
computer to solve a problem is known as an algorithm.
MEASUREMENT OF LENGTH,
PERIMETER AND AREA
Measurement of length, perimeter, and area is
a common task in GIS.
It is important to note that all measurements
from a GIS will be approximate due the reason
that vector data are made up of straight line
segments, and all raster entities are
approximated using a grid cell representation.
MEASUREMENT OF LENGTH
In a raster GIS, the Euclidean distance or the shortest path
between two points A and B can be measured by any one of
the methods given below.
(i)
By drawing a straight line between A and
=
B, and computing its length as hypotenuse
of a the right angled triangle ABC by
Pythagorean geometry
B
4
4.5
3.6
= 5.7 units.
42 42
4.2
2.8
1
A
3.6
4.5
c
2
5.0
e
5.7
g
i.e. AB = AC 2 CB 2 =
5.0
MEASUREMENT OF LENGTH
(ii) By measuring distances along raster
cell sides Aa, ab, bc, .. gB, and
adding up them, i.e.,
AB = (Aa)+(ab)+ . + (fg)+(gB),
or = 1+1+1+1+1+1+1+1 = 8 units
B
4
4.5
5.0
p
3.6
4.2
2.8
5.0
q
e
3.6
4.5
b
1
A
5.7
2
a
(iii)
By forming concentric equidistant zones around the
starting point A. Thus,
AB = 42 42 = 5.66 5.7 units
AP = 22 32 = 3.61 3.6 units
AQ = 42 22 = 4.47 4.5 units
MEASUREMENT OF PERIMETER
In the raster GIS, the area and perimeter measurements are
affected by location of origin and orientation of the raster
grid, and these problems are solved by proper selection of
grids with north-south alignment and use of consistent
origins.
Measurement of length, perimeter, and area in vector GIS is
easier and accurate compared to raster GIS. The length of
the line AB is calculated as below:
AB ( xb xa ) ( yb y a )
2
where (xa, ya) and (xb, yb) are the coordinates of A and B,
respectively.
MEASUREMENT OF AREA
The area of a feature is
calculated by totaling the areas
of simple geometric figures
formed by subdividing the
feature or directly by the
following formula
D (xd, yd)
A (xa, ya)
B (xb, yb)
O
Area of ABCDEA =
E (xe, ye)
C (xc, yc)
X
1
xa ( yb ye ) xb ( yc ya ) xc ( yd yb ) xd ( ye yc ) xe ( ya yd )
2
QUERIES
(i)
(ii) What is the route that will take minimum time to travel between two points
?
(iii)
QUERIES
QUERIES
The queries, which require spatial analysis, fall
under the class of spatial query, while aspatial
queries use only attribute data of features involving
no spatial analysis.
A query such as how many hospitals with heart care
facility are located in a given area is an aspatial
query as it can be performed by database software
alone and it does not involve analysis of the spatial
component of data.
RECLASSIFICATION
Reclassification is another form of query in raster
GIS with only difference that it results in a new
image in which the features of different classes have
different codes.
If cell value of a particular class of feature selected
is say 10, then in reclassification all the cells with
value 10 may be assigned a new value, say 1, and
the remaining cells with various classes of features
may be assigned a value, say 0.
RECLASSIFICATION
11 11 11 11 11
11 11 10 10 10
11 11 10 10 10 11 11 11 10 10
12 12 12 10 10 10 11 11 11 10
12 12 12 12 10 11 11 11 11 10
12 8
10 11 11 11 10
9
11 11 11
12 12 10 10 10 10 9
11 11 11
12 12 10 10 10 10 10 11 11 11
12 12 10 10 10 10 10 11 11 11
12
Wetland
Water
Agriculture
10
Forest
11
BUFFERING &
NEIGHBOURHOOD FUNCTIONS
Buffering function, also known as proximity function, in
GIS is one of the neighbourhood functions and it is used
to create a zone of interest around an entity, or set of
entities.
Buffering allows a spatial entity to influence its
neighbours, or the neighbours to influence the
character of an entity.
Other neighbourhood function that include data filtering,
involves the recalculation of cells in a raster image
based on characteristics of neighbours. If a point is
buffered, a circular zone is created (Fig. a), and buffering
lines and areas creates new areas (Fig.b and c).
r1
r2
d1
d
d1 d2
d2
CRITERIA
In order to conduct Point Pattern Analysis,
your data must meet five important criteria:
1. The pattern must be mapped on a plane,
meaning that you will need both latitude and
longitude coordinates.
2. A study area must be selected and
determined prior to the analysis.
3. The Point Data should not be a selected
sample, but rather the entire set of data you
seek to analyze.
4.
TECHNIQUES
1.) Quadrant Count Methods,
2.) Kernel Density Estimation (sometimes called KMeans), and
3.) Nearest Neighbor Distance
Clustered pattern
If
VTMR=1,
the
pattern is random.
This implies the data
set has no dominant
trend
towards
clustering
or
dispersion. A random
pattern may look like
as shown in Figure.
Random Pattern
ADVANTAGES
Very good for analyzing the point patterns to
discover the Hot Spots.
Provides with a useful link to geographical
data because it is able to transform the data
into a density surface.
Choice of r, the kernel bandwidth strongly
affects our density surface.
We can weight these patterns with other data
such as density of populations and
unemployment rates.
STEP 1
Compute the average of nearest neighbour
distance of the point pattern using the eq:Ad = (di)/n
Where
di is the distance from point i to its nearest
neighbour
n is the total number of points in the chosen
map area.
STEP 2
Compute the expected value of the average
nearest distance
Ed=1/2sqr(A/n)
where
A denotes the map area
RANDOM
UNIFORM
CLUSTERED
FILTERING
Filtering is one of the functions of
neighbourhood, used for processing of remote
sensing data.
Filtering changes the value of a cell, which
depends on the attributes of the cells in
neighbourhood.
A filter comprises of a group of cells around a
target cell, and its size and shape are decided
by the operator.
Mean filter
Weighted mean filter
Median filter
Mode filter
Olympic filter
MEAN FILTER
In this technique, each pixel within
a given window (say 33) is
sequentially examined, and, if the
magnitude of the central pixel is
significantly different than the
average
brightness
of
its
surrounding pixels (as given by a
predetermined threshold value),
then the central pixel is replaced
by the average value
a1
a2
a3
a4
a5
a6
a7
a8
X1
1
X
8
a
threshold value)
(given
i 1
i 1
X=X
where
X = the brightness value of the central pixel
ai = brightness value of the surrounding ith
pixel.
MEDIAN FILTER
Median filter uses median rather than average of
the neighborhood pixels of a given window.
The median of a set of numbers is that value such
that 50% of the number are above and 50% are
below it.
It is considered to be superior than the mean filter
primarily due to two reasons.
First of all, the median of a set of n numbers
(where n is odd) is always equal to one of the
value present in the data set. Second, median is
less sensitive to errors or to extreme data values.
MODE FILTER
In this technique, the central pixel is replaced
by its most common neighbour.
This is particularly useful in coded images
such as classification maps in which the pixel
values represent object labels.
Averaging labels makes no sense, but mode
filters may clean up the isolated points.
It produces irregular shifts in edges which
make them appear ragged.
OLYMPIC FILTER
The Olympic filter is a variant of a mean filter.
It is named after the system of scoring used in
certain Olympic events, where the highest and
lowest scores are dropped and the remaining
ones averaged.
The Olympic filter ranks the values within the
filter window (number of values = N), and
discards high and low values before
calculating the mean of those remaining.
The output of the Olympic filter is less
influenced by outlier values than the mean
filter, but the averaging process blurs edges
and removes fine detail from the images.
FILTERING
a
d
4
b
4
Filter
Target cell
FILTERING
The recalculated value of the target cell dc depends
upon the criteria used as given in Table.
Filtering may be required in the raster data obtained from
a classified satellite images to smoothen the noise
present in the data due to high spatial variability in a
particular class of feature such as vegetation cover, or
due to the problems with the data collection devices.
Criteria
Target cell
Original Value
New Value
Minimum filter
dc
Maximum
dc
Mean
dc
2.67
Mode
dc
Diversity
dc
TEXTURE TRANSFORMATIONS
Texture transformations are another set of
tools that are used to identify the spatial
pattern in data.
Texture filters are designed to enhance the
heterogeneity of an area.
A common algorithm for a texture filter is to
calculate the standard deviation of the cell
values in a 3 3 neighbourhood.
SLOPE TRANSFORMATION
A slope transformation turns a data layer of
elevation into one of slope, by calculating the
local first derivative. A companion to slope is
aspect. An aspect calculation is used to
determine the direction that a slope faces.
A mathematical way of explaining aspect is to
calculate the horizontal component of the vector
perpendicular to the surface. Aspect is usually
classified into bins of fixed size, so that the
resulting data layer is not continuous, but
ordinal.
A frequent choice is to classify slope into eight
categories, each representing an eighth of a
circle (or 45) range in aspect.
THANK
YOU
MAP OVERLAY
Integrating two or more different thematic map
layers of the same geographic area is a
common operation in GIS analysis.
The technique of map overlay has many
applications, such as for visual comparison by
overlaying a map showing only hospitals on a
road network map, to answer the query that
where are the hospitals located.
In this case no new data are produced. This
technique is also used for the overlay of vector
data on a raster background image, which is a
scanned topographic map.
VECTOR OVERLAY
In a vector-based system, the analysis is based
on a polygon intersection algorithm in which
new polygons are created as needed, and
redundant boundaries are eliminated (Fig.).
Vector map overlay relies heavily on the two
associated disciplines of geometry and topology.
The data layers being overlaid need to be
topologically correct boundaries so that the lines
meet at nodes and polygon boundaries are
closed.
To create topology for a new data layer produced
as a result of the overlay process, the
intersection of lines and polygons from the input
layers need to be calculated using geometry.
Residential
area
Class-B
residences
Business
centre
Residential
area
Commercial
area
Residential
area
=
Commercial
area
VECTOR OVERLAY
(i) Point-in-polygon,
(ii) Line-in-polygon, and
(iii) Polygon-on-polygon.
POINT-IN-POLYGON
POINT-IN-POLYGON
Point map
Polygon map
Point map
Sand
+
1
+
2
+
+
=
Clay
3+
+
+
Well
(a) Point-in-polygon
LINE-IN-POLYGON
To answer the queries like whether a road lies
on sandy or clay type of soil, line-in-polygon
overlay is used.
Fig. shows data layers of roads and soil types.
When the two layers are integrated, the roads
are split into smaller segments, depending on
which part of a road falls in which type of soil
category.
A database record of for each new road
segment is created in the output map. The
output layer is more complicated then the two
input layers as topological information is
required to be retained, and therefore, line-inpolygon is more complex.
LINE-IN-POLYGON
Line map
Polygon map
Line map
Sand
Clay
Road
1
2
3
4
5
Sand
Clay
Sand
Clay
Clay
(b) Line-in-polygon
POLYGON-ON-POLYGON
Polygon map
Sand
+
Clay
Urban
=
Union
Sand
Clay
Urban
=
Erase
Sand
Clay
Urban
(c) Polygon-on-polygon
Urban area on
sandy soil
=
Intersect (AND)
RASTER OVERLAY
In a raster-based system, cells in input data represent
the raster data structure.
A single cell represents a point, a string of cells
represents a line and a group of cells represents an
area.
Raster overlays employ mathematical operations of
addition, subtraction, multiplication or division on the
individual cell values of the input layers to produce
output data.
This requires appropriate coding of features
represented by points, lines, and areas the input data
layers.
RASTER OVERLAY
For example, wells are represented as 1 in
the well map layer while sewer lines are
expressed as 2 in the sewer line map layer.
In the land use map layer, the coding may be
3 for wheat field, 4 for forest, 5 for clay soil
and 6 for urban areas while for all the cells
having features of no interest as 0, in all data
layers.
If the codes assigned to different land uses in
different layers are same, the interpretation of
results becomes different.
Well
Wheat
Output map
Wheat field
Well
0
0
1
0
3
3
3
0
0
0
Well map
Well outside
wheat field
Output map
(a)
Urban map
Output map
Sewer line
Urban
Urban map
Output map
Soil map
Forest map
Output map
Clay soil
Forest
Soil map
4
4
4
4
4
0
0
0
Forest map
Output map
(c)
Operation: Addition
(i) Clay soil = 1
Soil map
Forest map
Forest
0
Clay soil
Soil map
Forest or clay
soil
(d)
Operation: Multiplication
(i) Clay soil = 1
Soil map
Forest map
Output map
Forest
Clay soil
Other areas
Soil map
Forest map
Output map
(e)
Ranking of qualitative
judgments
Intensity of
Importance
1
2
3
4
5
6
7
8
9
Qualitative Definition
Equal importance
Equal to moderate
Moderate importance
Moderate to strong
Strong importance
Strong to very strong
Very strong
Very to extremely strong importance
Extreme importance
1/7
1/5
1/7
1/9
1/3
1/7
1/9
1/5
1/3
1/3
Computation of criterion
weights
Step I
Compute the sum of each column,
Step II
Divide each entry in the matrix by its column sum
(Normalization)
Step III
Compute the average of the elements in each row of the
normalized matrix, i.e.,divide the sum of normalized scores for
each row by the number of criteria.
Step IV
Check to make sure the decision maker is consistent in making
the comparisons.
Consistency measurement
Step I
Determine the weighted sum vector by
multiplying the corresponding weight times the
original pairwise comparison matrix.
Step II
Consistency Index
= Consistency vector
N
CI
N 1
Consistency of Randomness
CI
CR
CR<0.1 forRI
consistency
CRITERIA Rainfall
SW
NDVI Tmax
RH
Rainfall
1.00
2.00
3.00
5.00
6.00
SW
0.50
1.00
2.00
3.00
5.00
NDVI
0.33
0.50
1.00
2.00
3.00
Tmax
0.20
0.33
0.50
1.00
2.00
RH
0.17
0.20
0.33
0.50
1.00
2.20
4.03
STEP II
CRITERIA Rainfall SW
NDVI Tmax
RH
WT
Rainfall
0.45
0.50
0.44 0.43
0.35
0.44
SW
0.23
0.25
0.29 0.26
0.29
0.26
NDVI
0.15
0.12
0.15 0.17
0.18
0.15
Tmax
0.09
0.08
0.07 0.09
0.12
0.09
RH
0.08
0.05
0.05 0.04
0.06
0.06
Consistency Check
STEP I
Rainfall
0.435 0.529
0.463
SW
0.218 0.265
0.309
NDVI
0.145 0.132
0.154
Tmax
0.087 0.088
0.077
RH
0.073 0.053
0.051
0.45
1
0.27
1
0.18
1
0.09
0
0.04
5
0.332
0.276
0.166
0.111
0.055
Consistency Check
STEP II
2.211 5.078
1.338 5.059
0.778 5.039
0.453 5.022
0.277 5.017
CI
RI
CR
0.011
1.120
0.010
5.043
SPATIAL AUTOCORRELATION
Spatial autocorrelation may be defined as the
relationship among values of a single variable that
comes from the geographic arrangement of the
areas in which these values occur.
It measures
the similarity of objects within an area,
the degree to which a spatial phenomenon is correlated
to itself in space,
the level of interdependence between the variables,
the nature and strength of the interdependence, i.e.
spatial autocorrelation is an assessment of the
correlation of a variable in reference to spatial location of
the variable.
DESCRIBING SPATIAL
AUTOCORRELATION
The general method of describing autocorrelation
in a variable is to compute some index of
covariance for a series of lag distances (or
distance classes) from each point.
The
resulting
correlogram
illustrates
autocorrelation at each lag distance.
Membership in a given distance class is defined
by assigning a weight to each pair of points in the
analysis; typically this weight is a simple indicator
function, taking on a value of 1 if within the
distance class, else 0
Weights may also be defined in other ways, in
which case they take on non integer values.
Geary's C
This statistic was developed by Roy C. Geary
Where:
N = the number of observations
Wij = the distance (spatial lag) between pixels
i and j
xi = the value at location i
xj = the value at location j
MORAN'S I
Measures the correlation (simultaneous
change in value of two numerically valued
random variables) among the neighboring
observations in the pattern is done using the
spatial autocorrelation statistic Moran's I.
Moran's I is defined as a measure of the
correlation among neighboring observations
in a pattern.
where
N is the number of spatial units indexed by i and j;
X is the variable of interest;
NETWORK ANALYSIS
In the context of GIS, a network is defined as
a set of interconnected linear features through
which resources can flow.
Common examples of networks include
highways,
- railways,
city streets,
- canals,
rivers,
transportation routes, such as
garbage collection, mail delivery, school buses, and
utility distribution systems, for example,
telephone, electricity, water supply, and sewage.
ROUTE TRACING
Route tracing through the network analysis is
required
to
identify
the
routes
for
unidirectional flow of resources with special
reference to stream networks and services,
such as sewerage systems and cable TV
networks.
Other applications are determination of
streams contributing to a reservoir, customers
serviced by a particular sewer main, or clients
affected by a broken cable.
Location-Allocation Modeling
An important application of network analysis is
allocation of resources.
This is done by modeling of supply and demand
which help of movement of goods, people, and
information or services through the network are
required to match the demand with the supply.
Allocation of resources is usually done by
allocating links in the network to the nearest
center of supply taking into account impedance
values.
The maximum catchment area of a particular
supply centre can also be determined on the
basis of the demand located along adjacent links
in the network.
Location-Allocation Modeling
A situation may arise by imposing limitation on
supply and demand, in which some parts of the
network may not be serviced despite a demand
being present in that part.
Such problems can be solved by reducing the
supply to some parts of the network or by
identifying the optimum location for a new centre
to meet the shortfall in supply relative to demand.
THANK
YOU