Raster Data Structures

Based on column/row structure, intersections called 'cells'

Examples of datasets commonly represented using the raster data structure

Satellite imagery

DEM - Digital Elevation Model, each cell represents an elevation value

Can convert any vector data to raster data to vector and vice-versa but there is always some error introduced when you do so

Spatial scale and cell size - basis for representativeness of the real-world in raster datasets

• Large cell sizes aggregate data
  • Smaller files, shorter processing time
  • Smoother boundaries (better for maps)
• Small cell sizes
  • Larger files, longer processing time
  • Stair-step boundaries
• Reducing the cell size by half (e.g. 20m to 10m) quadruples the number of cells

Choice of cell size is a function of

• Spatial variability of the feature in question
• Purpose/objective of the GIS
• Resolution of the other datasets

Examples - DEMs of 20m and 100m for Northern Indiana vs. Southern Indiana

Important Terminology:

"What is the cell size of the dataset?"

"What is the resolution of the dataset"

"What scale are the data?"

Minimum mapping unit

Cell assignment rules (database construction and data conversion (e.g. vector to raster)

Making cell assignments is relatively easy with homogenous areas. Heterogeneous areas (mixed areas) are more complicated. A number of cell assignments rules follow:

• Maximum area
• Center point
• Weighted Area
• Presence/absence - any area

Converting a line feature to a raster dataset using different assignment rules

Errors of omission and commission

• Error associated with misrepresentation of a dataset (applies to raster and vector)
• With raster datasets in particular refers to error within a cell related to cells that have a mixture of features but must be represented as a single value in the raster dataset
• Examples
  • Raster dataset of landcover at 10m, 50m, 100m
  • Converting data between resolutions, scaling up

How would the following areas be coded?

With raster data analysis it is common to integrate data from multiple sources

Important to be aware of the cell size of source data and impact on resulting analyses

In most cases you should convert your data to a common cell size, usually the coarsest cell size of all source datasets.

Application example one:

Representation of population density at 10km spatial resolution vs. 1km cell size.

Application example two:

Goal: integrate/overlay spatial datasets for population density (500m spatial resolution) and ozone concentration (100m spatial resolution) around Indianapolis. Explore the relationship between these two datasets.

Application example three:

Goal: integrate/overlay spatial datasets for population density (1km resolution) and one for hydrography (rivers, streams, lakes - ). Explore the correlation between these two datasets. How does the resolution of the hydrography affect the analysis? What are the implications of the raster cell assignment rule?

Resampling - Up-scaling and down-scaling

Possible to convert a raster dataset at one resolution to a raster dataset at another resolution. If cell resolutions are divisible then conversion is straightforward. If cell resolutions are not divisible, or if converting to a raster dataset in another geometric orientation then a resampling method must be used:

• Nearest Neighbor
• Bilinear Interpolation
• Cubic Convolution

Upscaling - producing a new derived dataset with a larger cell size/cell resolution

Aggregates data based on different in cell size

Most frequent value (nominal data), or interpolation of neighborhood of values (ratio data)

Downscaling - producing a new derived dataset with a smaller cell size/cell resolution

IT IS IMPOSSIBLE TO CREATE INFORMATION BY DOWN-SCALING!!

Simple cell assignment based on source cell (coarser resolution)

Why are raster datasets used to represent phenomena that vary continuously in the real world?

Data compression techniques reduce the data storage requirements for raster datasets in some situations

• The more homogenous a dataset the more effective the data compression
  • See here and here

Advantages/Disadvantages of Raster and Vector datasets

See Bolstad Chapter 2, Table 2-2

Characteristic	Raster	Vector
Data Structure	Simple	Complex
Storage	Large, esp. for heterogeneous areas	Usually small(er)
Coordinate Conversion	May require resampling	Simple
Analysis	Easy, facilitated by simple grid overlay	Good for network analysis, more complicated for polygon overlay
Positional Precision	Function of cell size	Function of coordinate precision
Modeling (Accessibility)	Easy because of simple data structure	Often complex
Display/Output	Good for images, stair -step problem w/ linear and areal features depending on cell size	Usually pretty