Department of Urban and Regional Planning
Khulna University of Engineering & Technology (KUET), Khulna-9203, Bangladesh
URP 2281 GIS and Remote Sensing
Raster Analysis
Tanmoy Mazumder
Assistant Professor, Department of URP, KUET
Raster Analysis
Zonal statistics (land use within a watershed/ward)
Interpolation (surface from sampled points)
Map algebra
Change detection
Spatial modeling
Location suitability
Surface analysis (slope, aspect, curvature)
Terrain analysis (elevation, slope, aspect, visibility, etc.)
Hydrology (stream networks, drainage basins, flooding)
And many others
Raster Operations
Local
Use data in a single cell to calculate output
Neighborhood
Use data from a set of cells (usually adjacent) to
calculate output
Global
Use all data from a raster data layer
Local Functions
Map Algebra - Applying Mathematical Functions
Cell by cell combination of Unary Binary
raster data layers
Layer addition, subtraction,
multiplication
Many different applications
Raster Calculator and Map Algebra
Example:
Cell by cell
Precipitation
evaluation of -
mathematical Losses
functions (Evaporation,
Infiltration)
=
Runoff
Logical Operations – AND, OR, NOT
AND Typical Input
Values
requires both corresponding
True ≠0
input cells be True for a true False = 0
output
OR
assigns True if either input
is True
NOT Output Values
Simply reverses the True 1 = True
0 = False
and False values N = NoData
Logical Operations - Ordinal or Equality Comparisons
Logical Operation in Change Analysis
Where did land cover change between 2009 and 2010?
Reclassification
Assign output values
that depend on a
specific set of input
values
Raster Overlay
Combine features from two or more layers
Features = grid cells OR groups of grid
cells with the same values
Often restricted to categorical data
Continuous data result in too many output
combinations Solution: convert continuous
data to categorical data first
Each unique combination of inputs is
assigned an output identifier
Attributes are combined in output
Remember Clip
Reduce or extract
Clip
data from larger,
more complex
datasets. Clipping
can be done
using a polygon
input or a raster
template or
“binary mask”
Masked Extents in Analyses
An analysis mask limits processing to areas that fall within the mask
Locations outside of mask are assigned NoData
Mask can be a raster or vector dataset (converted to raster)
Use a concurrent mask! (more on this later)
Neighborhood Functions
Depend on the concept of a
“moving window”
Moving window defines the input
values for the function
Output is usually associated with
the cell at the center of the window
position
Window is swept across the raster
layer, usually from left to right and
top to bottom
Neighborhood
Function
Examples
Neighborhood
Function
Majority Filter
Raster Compatibility for Analysis
Depends on
Extent
Cell size
Origin
Orientation
Incompatible raster cell
sizes, extents,
orientations
and
cause
?
ambiguities when raster
layers are combined
?
Raster Types
Orthogonal (Same cell size Concurrent (Orthogonal
and aligned in both direction, and Share the same
same orientation but extent extent, Overly exactly )
can be different)
Common Concurrency Problems
All of these rasters have the same number of rows and columns, but are not
concurrent. To do map algebra (correctly), rasters need to be compatible
(concurrent)
What should I do with non concurrent rasters
Resampling using environment settings to
ensure concurrency
Convert raster(s) to use the same extent,
origin, and cell size
1. Nearest neighbor assignment
2. Majority
3. Bilinear Interpolation
4. Cubic Convolution
Nearest Neighbor Assignment
Output raster contains
values from the nearest
neighboring cell in the
input raster
Does not alter values of
input
Best for discrete
(categorical) data
Bilinear Interpolation
Uses values of 4 nearest
input cell centers to determine
output value
New value is distance weighted
Results in a smoothed surface
Input values are modified
Better for continuous datasets
Cubic Convolution
Similar to bilinear but
weighted average is
calculated from the values
of the 16 nearest input
cells
Input values are modified
Interpolation - Estimate Values Between Known Values
A set of functions that predict values for a surface from a limited number of
sample points creating a continuous raster.
Nearest Neighbor
“Thiessen” Spline Interpolation
Polygon Interpolation
Zonal Statistics
Summarizing grids within zones
What is the average elevation with a watershed?
What is the area of each land cover class within a watershed?
Cartographic Modeling
Combining multiple layers to produce a new output
Cartographic Modeling
Where are suitable locations for a new park? Near a lake, close to a road, but avoid wetlands.
Thanks!!!