OE-6.4 GIS Concepts & Data Models

What this is and why it exists

Here is the definition worth carrying out of this unit. A geographic information system is a database in which every record knows where it is.

That join, between a shape on the ground and a row of attributes, is what makes it more than a map. A map shows you where things are. This lets you ask questions.

The recurring practical decision is raster against vector, and it has a clean rule behind it.

The vocabulary

• Spatial data — the geometry: where something is and what shape it has.
• Attribute data — the facts about it, as a table row.
• Join — the link between a geometry and its attributes.
• Raster model — the world as a grid of cells, each with a value.
• Vector model — the world as points, lines and polygons.
• Layer — one theme of data covering an area.
• Continuous field — a quantity defined everywhere, such as elevation.
• Discrete object — a thing with a boundary, such as a plot or a road.

The mental model

A system of this kind has the usual components: data, software, hardware, methods and people. The two most often underestimated are the last two. Bad method and untrained staff spoil good data reliably.

The join is the heart of it. A polygon on the map is connected to a row in a table. Selecting on the table selects on the map, and the other way round. Ask "which plots are larger than a hectare and zoned residential" and you get a map, not a list. That is the capability a paper map does not have and cannot be given.

Now the model choice, and it does have a clean rule. Raster for continuous fields, vector for discrete objects.

A raster is a grid of cells, each holding a value. That suits anything defined everywhere: elevation, temperature, rainfall, satellite imagery. Its resolution is the cell size, and everything inside a cell is one value — so detail below that size is gone, permanently.

A vector model stores points, lines and polygons with coordinates. That suits things with boundaries: parcels, roads, rivers, administrative areas. Boundaries stay exact at any zoom, and attributes attach naturally to each object.

Use the wrong one and the work becomes awkward rather than impossible. Storing a road network as a raster loses connectivity — you have coloured cells, not a network you can route along. Storing elevation as polygons means drawing contour bands and losing everything between them.

The applications follow from the join. Anything that combines *where* with *what* is a candidate, which is why the field turns up in agriculture, utilities, planning, disaster response and epidemiology alike.

What you should now be able to explain or do

Define such a system as a database where records know their location. Explain the join between spatial and attribute data and what it enables. Say which component is most often underestimated. Choose raster or vector from whether the thing is a continuous field or a discrete object. Say what is lost by storing a road network as a raster, and elevation as polygons.

Check yourself