UNIT 5 -
RASTER GIS CAPABILITIES
Compiled
with assistance from Micha Pazner,
University of Manitoba
1)
A. INTRODUCTION
2)
B. DISPLAYING LAYERS
a)
Basic
display
b) Other types of display
3)
C. LOCAL OPERATIONS
a)
Recoding
b) Overlaying layers
4)
D. OPERATIONS ON LOCAL NEIGHBORHOODS (FOCAL - Tomlin)
a)
Filtering
b) Slopes and aspects
5)
E. OPERATIONS ON EXTENDED NEIGHBORHOODS
a)
Distance
b) Buffer zones
c)
Visible
area or "viewshed"
6)
F. OPERATIONS ON ZONES (GROUPS OF PIXELS)
a)
Identifying
zones
b) Areas of zones
c)
Perimeter
of zones
d) Distance from zone boundary
e)
Shape
of zone
7)
G. COMMANDS TO DESCRIBE CONTENTS OF LAYERS
a)
One
layer
b) More than one layer
c)
Zones
on one layer
8)
H. ESSENTIAL
HOUSEKEEPING
9) REFERENCES
10) EXAM AND DISCUSSION QUESTIONS
NOTES
This
unit continues the overview of raster GIS. If possible, we suggest that you
replace and/or supplement the graphics provided with this unit with graphics
generated by the raster program your students will be using in their labs.
Alternatively, the best way to illustrate this unit may be through the use of a
laboratory demonstration.
Consider providing handouts to the students
that summarize the commands for the raster GIS program you will be using in
labs. Check your program's manual for a command summary or do a screen dump of
the appropriate help screen if there is one.
UNIT 5 - RASTER GIS
CAPABILITIES
Compiled
with assistance from Micha Pazner,
University of Manitoba
A. INTRODUCTION
A raster
GIS must have capabilities for:
- Input of data
- Various housekeeping functions
- Operations on layers, like those encountered in the
previous unit - recode, overlay and spread
- Integration with vector GIS operations
- Output of data and results
- The range of possible functions is enormous, current
raster GISs only scratch the surface
- Because the range is so large, some have tried to organize
functions into a consistent scheme, but no scheme has been widely accepted yet
- The unit covers a selection of the most useful and common
- Each raster GIS uses different names for the functions
IDRISI
is a commonly used and powerful raster based GIS developed by Dr. Eastman
at Clark University in Worcester, MASS. It is now a commercial success.
IDRISI TUTORIAL on line
at Univ. British Columbia
- ArcGIS Spatial Analyst provides
powerful tools for comprehensive, raster-based spatial modeling and analysis
- Find suitable locations
- Calculate the accumulated cost of traveling from one
point to another
- Perform land use analysis
- Predict fire risk
- Analyze transportation corridors
- Determine pollution levels
- Perform crop yield analysis
- Determine erosion potential
- Perform demographic analysis
- Conduct risk assessments
- Model and visualize crime patterns
-
- ESRI Clips
B. DISPLAYING
LAYERS
Basic display
- The simplest type of values to display are integers
- On a color display each integer value can be assigned a
unique color
- There must be as many colors as integers
- If the values have a natural order we will want the
sequence of colors to make sense
- E.g. elevation is often shown on a map using the sequence
blue-green-yellow-brown-white for increasing elevation
- There should be a legend explaining the meaning of each
color
- The system should generate the legend automatically based
on the descriptions of each value stored with the data layer
- On a dot matrix or laser printer shades of grey can be
generated by varying the density of dots
- If there are too many values for the number of colors, may
have to recode the layer before display
Other types of
display
- It may be appropriate to display the data as a surface
- Contours can be "threaded" through the pixels
along lines of constant value
- The searching operation for finding contours is
computer-intensive so may be slow
- The surface can be shown in an oblique, perspective view
- This can be done by drawing profiles across the raster
with each profile offset and hidden lines removed
- The surface might be colored using the values in a second
layer (a second layer can be "draped" over the surface defined by the
first layer)
- The result can be very effective
-
FLY Overs -- "LA The Movie" was produced by
Jet Propulsion Lab by draping a Landsat image of Los
Angeles over a layer of elevations, then simulating the view from a moving aircraft we've come along
way in 15 years - Google Earth and Microsoft's Virtual Earth - common
place.
-
These
operations are also computer-intensive because of the calculations necessary to
simulate perspective and remove hidden lines
C. LOCAL OPERATIONS
- Produce a new layer from one or more input layers
- The value of each new pixel is defined by the values of
the same pixel on the input layer(s)
- Neighboring or distant pixels have no effect
- Note: arithmetic operations make no sense unless the
values have appropriate scales of measurement (see Unit 6)
- You cannot find the "average" of soils types 3
and 5, nor is soil 5 "greater than" soil 3
Recoding
/ reclassing
- using only one
input layer
- Examples:
1. Assign a new value to each unique value on the input
layer
- Useful when the number of unique input values is small
2. Assign new values by assigning pixels to classes or
ranges based on their old values
- E.g. 0-499 becomes 1, 500-999 becomes 2, >1000 becomes
3
- Useful when the old layer has different values in each
cell, e.g. elevation or satellite images
3. Sort the unique values found on the input layer and
replace by the rank of the value
- E.g. 0, 1, 4, 6 on input layer become 1, 2, 3, 4
respectively
- Applications: assigning ranks to computed scores of
capability, suitability etc.
- Some systems allow a full range of mathematical operations
- E.g. newvalue = (2*oldvalue + 3) 2
Overlaying
layers
- An overlay occurs when the output value depends on two or
more input layers
- Many systems restrict overlay to two input layers only
- Examples:
1. Output value equals arithmetic average of input values
2. Output value equals the greatest (or least) of the input
values
3. Layers can be combined using arithmetic operations
- x and y are the input layers, z
is the output
- Some more examples:
Z = X + Y
Z = X * Y
Z = X / Y
4. Combination using logical conditions
- E.g. if y>0, then z = y ,
otherwise z = x
- Note: in many raster packages logical conditions cannot be
done directly from input layers
- must first create reclassified
input images so that cells have 0 if they do not meet the condition and 1 if
they do
Boolean logical operations on rasters (2 pages)
5. Assign a new value to every unique combination of input
values
-
E.g. LAYER 1 LAYER 2 OUTPUT LAYER
1 A 1
1 B 2
2 A 3
2
B 4
etc.
D. OPERATIONS ON
LOCAL NEIGHBORHOODS
- the value of a pixel on the new layer is determined by the
local neighborhood of the pixel on the old layer
Filtering
- A filter operates by moving
a "window" across the entire raster
- E.g. many windows are 3x3
cells
- The new value for the cell
at the middle of the window is a weighted average of the values in the window
- By changing the weights we
can produce two major effects:
- Smoothing -- a "low
pass" filter, removes or reduces local detail
- Edge enhancement -- a
"high pass" filter, exaggerates local detail
- Weights
should add to 1
- Example
filters:
1)
0.11 0.11 0.11
0.11
0.11 0.11
0.11
0.11 0.11
- Replaces
each value by the simple unweighted average of it and
its eight neighboring values
- Severely
smoothes the spatial variation on the layer
2)
0.05
0.05 0.05
0.05 0.60 0.05
0.05 0.05 0.05
- Gives
the pixel's old value 12 times the weight of its neighboring values
- Slightly
smoothes the layer
3)
-0.1 -0.1
-0.1
-0.1 1.8
-0.1
-0.1 -0.1
-0.1
- Slightly
enhances local detail by giving neighbors negative weights
Spatial
filtering
- Filters
can be useful in enhancing detail on images for input to GIS, or smoothing
layers to expose general trends
Three examples from IDRISI
G:\GEOG250\idrtutor\s_tools3.htm#lowpass
G:\GEOG250\idrtutor\s_tools3.htm#highpass
3.Directional filters
Enhance or detect directional structures in
the filtered images.
\\LAPIS\ELMES$\GEOG250\idrtutor\s_tools3.htm
- directional
Slopes
and aspects
- If the
values in a layer are elevations, we can compute the steepness of slopes by
looking at the difference between a pixel's value and those of its adjacent
neighbors
- The
direction of steepest slope, or the direction in which the surface is locally
"facing", is called its aspect
- Aspect
can be measured in degrees from North or by compass points - N, NE, E etc.
(Cyclic level of measurement)
- Slope
and aspect are useful in analyzing vegetation patterns, computing energy
balances and modeling erosion or runoff
- Aspect
determines the direction of runoff
-
This can be used to sketch drainage paths for runoff
Computing
Slope and Aspect in IDRISI with SURFACE
http://gis01.ame.umontreal.ca/APA/6237/idrtutor/s_tools4.htm
E. OPERATIONS ON
EXTENDED NEIGHBORHOODS
Distance
-
calculate the distance of each cell from a cell or the nearest of several cells
- each pixel's value in the new layer is its distance from the
given cell(s)
- Buffers
around objects and features are very useful GIS capabilities
- E.g.
build a logging buffer 500 m wide around all lakes and watercourses
- Buffer
operations can be visualized as spreading the object spatially by a given
distance
- The
result could be a layer with values:
1 if in original selected object
2 if in buffer
0 if outside object and buffer
- Applications
include noise buffers around roads, safety buffers around hazardous facilities
-
in many programs the buffer operation requires the user to first do a distance
operation, then a reclassification of the distance layer
- The
rate of spreading may be modified by another layer representing
"friction"
- E.g.
the friction layer could represent varying cost of travel
- This
will affect the width of the buffer - narrow in areas of high friction, etc.
Visible
area or "viewshed"
- Given a layer of elevations, and one or more viewpoints,
compute the area visible from at least one viewpoint
- E.g. value = 1 if visible, 0 if not
- useful for planning locations of
unsightly facilities such as smokestacks, or surveillance facilities such as
fire towers, or transmission facilities
F. OPERATIONS ON
ZONES (GROUPS OF PIXELS)
Identifying
zones
-
By
comparing adjacent pixels, identify all patches or zones having the same value
- Give each such patch or zone a unique number
- Set each pixel's value to the number of its patch or zone
Areas
of zones
- Measure the area of each zone and assign this value to
each pixel instead of the zone's number
- Alternatively output may be in the form of a summary table
sent to the printer or a file
Perimeter of zones
- Measure the perimeter of each zone and assign this value
to each pixel instead of the zone's number
- Alternatively output may be in the form of a summary table
sent to the printer or a file
- Length of perimeter is determined by summing the number of
exterior cell edges in each zone
- Note: the values calculated in both area and perimeter are
highly dependent upon the orientation of objects (zones) with respect to the
orientation of the grid
Overhead - Area and perimeter functions in rasters
- However, if boundaries in the study area do not have a
dominant orientation such errors may cancel out
Distance
from zone boundary
- Measure the distance from each pixel to the nearest part
of its zone boundary, and assign this value to the pixel
- Boundary is defined as the pixels which are adjacent to
pixels of different values
Shape
of zone
- Measure the shape of the zone and assign this to each
pixel in the zone
- One of the most common ways to measure shape is by
comparing the perimeter length of a zone to the square root of its area
- by dividing this number by 3.54 we get a measure which
ranges from 1 for a circle (the most compact shape possible) to 1.13 for a
square to large numbers for long, thin, wiggly zones
- Commands like this are important in landscape ecology
- Helpful in studying the effects of geometry and spatial
arrangement of habitat
- E.g. size and shape of woodlots on
the animal species they can sustain
- E.g. value of
linear park corridors across urban areas in allowing migration of animal
species
G. COMMANDS TO
DESCRIBE CONTENTS OF LAYERS
- Important to have ways of describing a layer's contents
- Particularly new layers created by GIS operations
- Particularly in generating results of analysis
One
layer
- generate statistics on a layer
- e.g. mean, median, most common
value, other statistics
More
than one layer
- Compare two maps statistically
- E.g. is pattern on one map related to pattern on the
other?
- E.g. chi-square test, regression, analysis of variance
Zones
on one layer
- Generate statistics for the zones on a layer
- E.g. largest, smallest, number, mean area
H. ESSENTIAL
HOUSEKEEPING
- List available layers
- Input, copy, rename layers
- Import and export layers to and from other systems
- Other raster GIS
- Input of images from remote sensing system
- Other types of GIS
- Identify resolution, orientation
- "Resample"
- Changing cell size, orientation, portion of raster to
analyze
- Change colors
- Provide help to the user
- Exit from the GIS (the most important command of all!)
CARTOGRAPHIC
MODELING EXAMPLE
Harvard Graduate
School of Design
REFERENCES
Berry,
J.K., 1987. "Fundamental operations in computer-assisted
map analysis," International Journal of Geographical Information
Systems 1:119-136. Describes a logical and
consistent way of classifying and grouping raster GIS functions.
Burrough, P.A., 1986. Principles of
Geographical Information Systems for Land Resource Assessment, Clarendon,
Oxford. Chapter 5 is a comprehensive review of raster GIS.
Star, J.L. and J.E. Estes,
1990. Geographic Information
Systems: An Introduction, Prentice Hall. A comprehensive
text on GIS, with excellent treatment of raster systems.
Tomlin, C.D., 1990. Geographic
Information Systems and Cartographic Modeling, Prentice-Hall, Englewood
Cliffs, NJ. A comprehensive approach to analysis and
modeling using raster systems - an excellent introduction to GIS-based
analysis.
User documentation for any raster GIS.
EXAM AND DISCUSSION
QUESTIONS
1.
Discuss the classification scheme proposed by Berry in the article listed in
the references. Is it logical and comprehensive? Can you suggest improvements
based on the material in this unit or the functions of a specific raster GIS to
which you have access?
2. A variety of user interfaces have been
used in raster GISs, including typed commands, menus and responses to
questions. Discuss the advantages and disadvantages of each.
3. "The most valuable skill in GIS is
the ability to take a real problem and convert it into a series of GIS
operations". Discuss.