Chapter 6

Managing raster data

(Mastering ArcGIS Pro by
Maribeth Price)
The raster model
Source: USGS
A raster stores an n×m array of
values in cells or pixels
representing squares on the
ground
The raster is georeferenced to an
earth location by an x-y in the
corner and a specified cell size
This raster stores elevation values
shaded from low (blue) to high
(white)

2
 Examples of rasters
Sources: USGS (a, c, e); Black Hills National Forest
(d); Esri (b,f) a) Digital Elevation Model
(DEM)
b) color aerial photograph
c) land use categories
d) primary, secondary,
and primitive roads
e) scanned topo map, or
Digital Raster Graphic
(DRG)
f) ArcGIS Online basemap
tile

3
Raster data structure
➢Dividing space into discrete uniform units-square cells,
namely a cellular model of geometry.
➢Location is inherent in the storage structure, namely,
implied by row and column number of the grid cell
rather than through the use of explicit spatial
coordinates.
➢Raster data structure does not provide precise
locational information.
➢Two-dimensional array of grid cells is called a layer, a
grid, or a band in different contexts.
➢Each layer of raster data is often used to represent a
particular topic (theme).

6-4
 Spatial resolution of raster
data
➢The cell must be small enough to capture the
required spatial detail.
➢Resampling to large cell size (low resolution) will
cause losing some of the small spatial features. But
down-sampling from large cell size to small cell size
won’t increase the data resolution!
➢Reducing the grid cell size to half the current size
will increase the data volume four times.
➢Store one number for every sq m, 1 Pb (trillion
bytes) www.ccsf.caltech.edu/~roy/dataquan/

(Image from Wikipedia)

6-5
Grid Cell/Square based Raster
Model
Columns

Grid
cell
Rows

Resolution

6-6
Pixels/Cells
• Each pixel contains one
numeric value
• The dimensions of a pixel is
the resolution
• Value represents some
property of that pixel area,
e.g. elevation or rainfall
• Values may be integers or
floating point numbers
• Unlike a polygon, each cell has
only ONE attribute; storing
multiple values requires
storing multiple rasters
© 2020 McGraw-Hill Education. 6-7
Raster resolution
Source: Black Hills National Forest

The x and y dimensions of each pixel define

the resolution of a raster
The precision of a raster is limited by the
resolution
Storage requirements increase by the square
of the resolution. This 10-meter raster of roads
takes up 9 × 9 or 81 times more space than 8
 General Rules for Selection of
Grid Cell Size (resolution)
Whittaker-Shannon Sampling Theorem:
➢the cell size must be smaller than half of the
minimum feature (minimum map units) that
you intend to represent.

➢The commonly suggested cell size is 1/5 -

1/7 of the minimum feature to be captured.

6-9
6-10
Pixel depth
Pixel depth is the number of bytes used for each pixel
• More bytes = larger numbers = more space.
Integer values
• 8-bit pixel (one byte) stores 28 values from 0 to 255.
• 16-bit pixel (two bytes) stores 216 values from 0 to 65,565.
• 24-bit pixel (three bytes) stores 224 values from 0 to 16.7
million.
Floating point values
• Required for decimal number storage.
• Uses a 32-bit pixel (four bytes).
• Includes the mantissa (significant figures) and an exponent.
• 1.2345678 × 103.
6-11
Change the raster pixel depth and
type
• Use "Copy Raster"
• Select a pixel type from the
list
• The new raster data set will
have the defined type
• Warning: converting a
floating point type to an
integer type will cause loss
in the precision of raster
values.
• For example, 2.533 -> 2 will
lose three significant digits.

12
Bands

• A band is a single raster

• Multiband rasters can contain multiple raster arrays
with the same number of rows and columns and the
same resolution, stored together in a multiband stack
• Satellite images and color aerial photographs are
common examples of multiband rasters
• This Landsat image of Crater Lake, OR contains seven
bands of brightness values in different wavelengths
6-13
 Displaying multiband rasters
Source: USGS

Computer displays mix red, green, and blue light to form

colors on the screen
A raster band can be assigned to any of the three colors
Some band combinations for this Landsat image include
a) Natural color (red, green, blue wavelengths)
b) False color (near infrared, red, blue)
c) Band 7, near infrared, blue
6-14
Continuous rasters
Sources: USGS (a); Esri (b)
Continuous rasters store
numeric values that can be
measured anywhere, such as
elevation (a), temperature, or
precipitation
Imagery (b) is a type of
continuous raster that stores
brightness (reflectance)
• Brightness is typically
stored as a digital number
(DN) between 0 and 255.

15
Discrete rasters
Sources: USGS (c); Black Hills National Forest (d)
• Discrete rasters store categorical
data such as land use (c), or
features such as roads (d)
• Many discrete rasters are
converted from vector points,
lines, or polygons
• Unlike vector data that can store
many attributes for each feature,
discrete rasters can only store
one attribute
• Digital raster graphic (DRG) is
the scanned product of US
Geological Survey topographic
maps.

16
Raster pyramids
Source: Esri
• Pyramids may be
built for a raster to
speed its display
• Successively lower
resolution copies are
created by
representing four
adjacent cells as a
single pixel
• Each copy is efficient
for different scales of
display
• Building the
pyramids takes a
6-17
little time and
Resampling

• Resampling is changing the resolution or

location of a raster, producing a new copy
• In this example, the original 30m resolution
grid (color cells) is being converted to a 40m
resolution grid shown by the black outline
cells
• A rule must be chosen to determine the
6-18
Why resample?

• Resampling is required when projecting rasters or when

analyzing two rasters with different resolutions (note: use
Project Raster in ArcGIS to project rasters)
• Resampling to a coarser resolution degrades the
precision and accuracy of a raster
• Resampling to a finer resolution does NOT improve the
data quality because you are simply using more cells to
store the same information

6-19
Resampling methods (2 of 2)

Nearest neighbor resampling grabs the

value from the old cell that falls at the
center of the new cell. It preserves the
original value and should always be used
with categorical data, or when the original
data values need to be preserved. It is the
fastest method
Bilinear resampling calculates a new
value from the four cells that fall closest to
the center of the new cell. It uses a
distance-weighted algorithm based on the
old cell centers. It is best used with
continuous data such as elevation
Cubic convolution resampling calculates
a new value from the sixteen cells that fall
closest to the center of the new cell. It uses
6-20
Resampling methods (1 of 2)
Source: Esri

Block resampling
groups pixels and
determines the new
value using a statistic
This pyramid uses block
resampling with the
majority statistic
Other rules such as
minimum, mean, or
maximum can also be
used

6-21
Raster tables Attribute tables are created
for discrete rasters with a
Source: Esri
limited number of unique
values
● The Value field contains
each unique value.
● The Count field
indicates how many cells
have that value.
● Explanatory fields, for
example the geology
units names from the
original polygon data set,
may also be included.
● Not all raster datasets
have a value attribute 6-22
NoData NoData is a special value
used in some raster formats
Source: USGS
to indicate absent data
Often replaced by zero in
other formats, but not as
satisfactory
• Zero is a valid precipitation
measurement, for example
(it didn’t rain).
Rasters are always
rectangular
• This DEM has NoData
values outside of South
Dakota.

6-23
BIP, BIL, and BSQ formats
• Multiband raster data can be stored in different formats
• Most common formats are JPEG, TIFF, BMP, Erdas
Imagine image
• ArcGIS has the File Geodatabase Raster (FGDR)
• Raw formats: Band Interleaved by Pixel (BIP), Band
Interleaved by Line (BIL), and Band Sequential (BSQ)

6-24
Header file for BIL, BIP, and BSQ
• BIL, BIP, and BSQ data are comprised of two files: an
ASCII header file (*.hdr) and a binary data file (*.bil,
*.bip, or *.bsq)
• Header files are provided with BIL, BIP, or BSQ files
• The file extension is .hdr
• Keywords are: nrows, ncols, bands, nbits, pixeltype,
byteorder, ulymap, ulxmap, xdim, ydim, bandrowbytes,
etc.
Sample BIL header file
Lines that don't begin with a keyword are treated as
comments.
nrows 1024 Comments can be placed here as well.
ncols 1024
nbands 3
6-25
nbits 8
Example of a BIL raster
nrows 8
ncols 8
nbands 3
layout bil

26
ASCII Raster
• Rasters can be presented in texts
• A header part shows the information of the raster
• Data part contains an array of values with each number
representing on pixel/cell of the raster
• Save as *.asc file
NCOLS xxx
NROWS xxx
XLLCENTER xxx
YLLCENTER xxx
CELLSIZE xxx
NODATA_VALUE xxx
row 1
row 2
...
row n
6-27
 Compression
• Many raster formats use compression to reduce storage size
• Run-length encoding (RLE) is used for discrete rasters and
stores how many times each adjacent value appears
• It is called reversible or lossless because the original can be
recovered.
• LZ77 is a generalized RLE
(0 1 1 1 1 1 3 3 3 3 3 3 5 5 5 5 5 5 7 8) to (01 15 36 56 71 81)

• Truncating decimals is another type of compression

• Truncation is irreversible, or lossy, because the original cannot
be recovered.
(1.2 3.6 4.2 9.8) to (1 3 4 9)

6-29
Raster coordinate systems
Source: USGS

• Rasters have coordinate

systems like other spatial
data sets
• A DEM in a GCS (a) has a
different shape and different
rows and columns than a
DEM in a conic coordinate
system
• Rasters are usually stored in
a projected coordinate
system (b) that preserves
area or distance to ensure
that analysis functions work
correctly 6-30
Projecting rasters • Projecting rasters is more
complex than projecting
Source: Esri
vector data and requires
resampling
• It takes more memory and
may introduce artifacts,
so it should be done as
few times as possible
• On the fly projection of a
map is most efficient
when the map CS
matches the raster CS

6-31
Georeferencing rasters
Source: Esri

Scanned maps have a CS in pixels and must be georeferenced

Users identify pairs of ground control points visible in the raster
and in a reference dataset
A transformation then calculates real-world coordinates from the
control point pairs and applies it to the raster
Affine transformations only adjust scale and rotation; polynomial
transformations allow differential adjustment, or rubbersheeting
X = a*x + b*y + c; Y = d*x + e * y + f
The root mean square error (RMSE) is an assessment of the error
of the transformation and should be recorded in the metadata
6-32
3D data
• 3D data include height or
Source: Esri
elevation (z values) in
addition to the x-y values
• Many different formats
are used to store 3D data
• An elevation source like a
DEM can be used to
drape 2D data on a
surface by aligning the x-
y locations
• 2D features with a base
height field, like
buildings, can be
extruded to form a 3D
representation
33
Digital Elevation Models
Source: USGS
• A Digital Elevation Model,
or DEM is a common 3D
raster that stores
elevation values
• These DEM examples are
symbolized to show the
lowest elevations in blue
and the highest in white
• DEMs are common forms
of 3D surface data, and
are used in many types of
analysis, such as creating
slope or aspect maps
6-34
 Sources of DEMs
• Global
• NASA's EarthExplorer
• USGS GTOPO30
• Resolutions (30m ~ 100m)
• For the US
• USGS 3DEP
• NOAA's digital coast
• Resolutions (1m - 30m)
• Technologies
• Photogrammetry
• LiDAR
• Interferometry SAR

35
Draping an image on an elevation
model
1. Insert a new local scene
2. Add the image to be
displayed
3. In "Elevation Surfaces",
right click and select create
surfaces from sources and
select the elevation data as
the source
4. Right click the image layer
and check the properties of
it.
5. In the elevation tab, set the
Elevation as "Features are
On custom elevation
surface"
6. Select custom surface as
the new surface you just
36
created from the source
Extruding a polygon data with a 3D
field
• Insert a new local scene
• Add the polygon data to the
map
• In Featurelayer ribbon and
Extrusion tab, select type
as " min height"
• Set field as an expression,
such as pop2014 / 10
• The polygons will be
extruded as 3D objects

37
Point clouds
Source: Esri

Point clouds store

millions of x,y,z triplets
LiDAR (Light Detection
and Ranging) produces
point cloud data sets
Often colored by
elevation to visualize
heights.
This scene shows the
river and ground in blue
and the bridge and
building heights in red
6-38
TINs
Source: USGS

• A triangular
irregular network
(TIN) stores x,y,z
triplets (nodes) that
define the corners of
triangular facets
bounded by linear
edges
• TINs efficiently store
3D surfaces by using
fewer nodes for flat
areas and more nodes
for steep areas

6-39
Space time cube
Source: Esri, USGS

A space time cube sieves

x,y points into a regular
grid and displays an
attribute through time
using height
This map shows the
summed magnitudes of
earthquakes in space and
time
Dark purple volumes show
where and when shaking
has been most intense

6-40
Creating a space time cube
Source: Esri A correctly formatted
Date/Time field is required
This data set contains
earthquakes over a 36 hour
period, with magnitudes, to be
aggregated in 6 hour time
steps
A rectangular or hexagonal
grid, with specified size, may
be selected
Choose the field(s) to be
aggregated in the grid/time
step interval
The standard netCDF data 6-41