Raster Analysis - Lecture Material - Completely GIS dan Remote Sensing tutorial - facegis.com
Raster Analysis

1. Introduction

Geographic data can be stored either in raster or vector format. Most data analysis can be performed either in raster or vector environments (although image data must be in raster). DEM data can be stored either as GRIDs or TINs (see DEM lecture). All point, line and polygon data can be stored in either raster or vector.

As an aside: All general computer files are of two types:
ASCII: easily read, edited, exchanged but bulky for graphics data
Binary: require specific software to read, more compact than ASCII.

GIS spatial data is usually (but not always) binary and are either raster or vector:

Vector data are based on features and have x and y coordinates
You are able to ask questions such as: What are the characteristics of this feature ?

Raster data are based on pixels, with a grid like system of rows and columns.
We can ask questions such as: What is at this location ?

2. Raster data model

For each layer, each grid cell (pixel) contains one numeric value, row-by-row:

present/absent = 0-1, or 0-255 (8 bit), or 0-65536 (16 bit), or float (decimals)

  • Simple 'grid' structure of rows and columns.
  • Based on cells or picture elements (pixels).
  • Linear feature (e.g. a road) is a contiguous sequence of cells.
  • cell value is based on a selected attribute
  • Resolution is based on size of cell -> the smaller the cell, the higher the resolution

Header information (which may be a separate file, or at the start of the data file); includes information on:

  • number of rows and columns.
  • X and Y coordinate of upper left and lower right corners.
  • pixel size usually in round values, e.g. 10, 25, 250, 1000 metres
  • georeferencing is 'implicit' (based on header information)


  • A simple data structure.
  • Overlay operations are straight forward (see figure below)
  • High spatial variability is efficiently represented (e.g. relief).
  • Only raster can easily store image data (e.g. photos).


  • Data structure is not compact (but relative disadvantage depends on layer complexity)
  • Limited capability in attribute management: each pixel usually has one data value for each layer
  • Map output can appear 'blocky'.

Example: .jpg , .tif (image) and .tfw (world file) ; or geotiff (georeferenced)

3. Vector Data model

  • Features are coded as points, lines (arcs) and areas (polygons).
  • Defined by single points, connected nodes, and arcs.
  • Vector files contain information attached to features.
  • georeferencing is explicit - coordinates on each point or vertex


  • Compact data structure for generally homogenous areas
  • Efficient encoding of topology (= containment, contiguity, connectivity)
  • Strong in database management
  • Better suited for map output.


  • More complex data structure.
  • Some types of analysis are more complex
  • Cannot store (continuously varying) image data.

Example: shapefiles - consists of at least three files: .shp .shx .dbf ( also for projections: .prj )

CAD files: .dgn (Microstation), .dxf (Autocad)

Further example of TIN (Vector) versus GRID DEM (raster) see DEM Lab and surface analysis

Raster is more efficient for overlay analysis

GIS and Remote Sensing tutorial of facegis.com
GIS and Remote Sensing tutorial of facegis.com

    [refer also back to this figure below in vector GIS analysis]
    Extracting specific info from a data layer and combining it with info from that layer or another layer can be related to Boolean algebra, expressions and operators. Examples of boolean logic using Boolean operators: (see below)

    GIS and Remote Sensing tutorial of facegis.com
    Maps A, B, and C represent the original map layers.
    The shade areas in D, E, F, and G represent where a condition is met for those locations.
    Map D shows : "Where is condition A but not B met?"
    e.g. Which areas have deep soils but NOT spruce?
    Map E shows : "Where are both conditions A and D met?"
    e.g. Which areas have deep soils AND spruce?
    Map F shows : "Where are either conditions B or C met?"
    e.g. Which areas have spruce OR good drainage?
    Map G shows : "Where is condition B or C, but not both, met?"
    e.g. Which areas have either spruce OR good drainage but NOT both?

4. Conversion Between Raster & Vector

Rasterisation: Vector -> Raster
  • Place grid cell over 'map': simple conversion.
  • Code whether a feature lies in a cell or not in a cell.
  • Simple process: one decision is the size of the cell to be produced
  • Decide on which attribute the new raster will store and display
Vectorisation: Raster -> Vector
  • 'Thread' a line through similar pixels: more complex process.
  • Use thinning and linking, requires editing.
  • Complex process: depends on available software coversion algorithm. ('R2V') and arcscan

5. General Types of Analysis in GIS

a. Database Query
- area, perimeter, values
b. Overlay
- compare different layers
c. Algebra
- modify by a given factor: add, subtract, multiply, divide graphic
d. Transform
- modify by projection, datum or geocorrect
e. Classify
- dissolve, group, merge, generalize
f. Distance
- cost surfaces, distance to features, buffering
g. Network
- hydrological, transportation, utilities, animal migration e.g. network analyst ** work opportunity (see overhead in class)
h. Statistics
- filtering, smoothing, 3D surfacing (surface analysis) avalanche prediction example
i. Modeling
- e.g. fire spread: based on fuel type, wind speed, direction, buffers, topography e.g. polar ice; pine beetle animation

SINMAP (slope stability) http://hydrology.neng.usu.edu/sinmap

SELES (Landscape Modelling) http://www.cs.sfu.ca/research/SEED/seles.htm

Other free raster downloads: http://www.for.gov.bc.ca/hre/dulp/download.htm

Hawth analysis tools for arcview in spatialecology: http://www.spatialecology.com

Raster GIS:

Grass: http://grass.itc.it/see applications and Vanderhoof viewshed analysis

Idrisi: www.clarklabs.org see applications

ArcGIS: see Spatial Analyst (formerly know as GRID)

6. Review

Things you should know after finishing this lecture:
  1. How do vector & raster systems differ in regards to strengths in database management ?
  2. In which types of analysis are raster data better
  3. why is vector to raster conversion easier than vice versa ?
  4. Are raster data always bulkier than vector - where is it most noticeable ?
  5. Is a line vector or raster ? (explain why)

Source: Source: http://www.gis.unbc.ca/courses/geog300/lectures/lect11/index.php