Scale, Accuracy, and Resolution in a GIS

Because GIS data is stored in a very different way than paper map data, the relationships between map scale, data accuracy, resolution, and density are very different between GIS and paper maps.

Map scale

Map scale specifies the amount of reduction between the real world and its graphic representation (usually a paper map). It is usually expressed as a ratio (eg 1:20,000), or equivalence (eg 1 mm = 20 m). Since a paper map is always the same size, its scale is fixed when it is printed, and cannot change.

However, a map in a GIS can be shrunk or enlarged at will on the screen or on paper. You can zoom in until the screen displays a square metre or less, or zoom out until the screen displays all of BC. This means that geographic data in a GIS doesn't really have a 'map scale'.

Display scale

The display scale of a map is the scale at which it 'looks right'. Because a paper map is created at certain scale, its 'map scale' and 'display scale' are the same.

The display scale influences two things about a map :

If you put a 1:20,000 scale paper map on an reducing photocopier, you can make it into a 1:100,000 map (ie reduce it by a factor of 5). However, probably areas of detail will be merged into big black blobs, and most of the text on the map will be too small to read.

A GIS map's annotation (ie text and symbols) must be designed with a display scale, just like a paper map. There is a range of scale in which it will 'look right', even though it is possible to display it at other scales with the GIS software.

Data accuracy and uncertainty

Data accuracy is an statement of how closely a bit of data represents the real world. It applies to geographical information in all these ways: This last, or 'locational' accuracy is of interest here, and is generally stated in terms of uncertainty. For example, '95% of the well locations are within 50 metres of their surveyed locations'.

A rigorous statement of accuracy will include statistical measures of uncertainty and variation, as well as how and when the information was collected. Spatial data accuracy is independent of map scale and display scale, and should be stated in ground measurement units.

Data resolution

Data resolution is the smallest difference between adjacent positions that can be recorded. Since a paper map is always the same size, its data resolution is tied to its scale. Resolution also limits the minimum size of feature that can be stored.

Generally, a line cannot be drawn much narrower than about 1/2 a millimetre. Therefore, on a 1:20,000 scale paper map, the minimum distance which can be represented (resolution) is about 10 metres. On a 1:250,000 scale paper map, the resolution is 125 metres.

However, most GIS store locations in ground units (eg UTM coordinates, or Longitude/Latitude) with a resolution of a centimetre or less. This resolution is far greater than the uncertainty of any of BC Environment's data.

Raster data resolution

Raster data is stored as (usually square) pixels, which form a grid or mesh over an area of the earth. The size of these pixels determines the resolution of the raster, because it is impossible to store anything which falls 'between' the pixels. A GIS allows raster pixels to be any size, although they should not be smaller than the uncertainty of the data.

If a raster coverage is derived from vector linework, its pixels should not be smaller than the uncertainty in the linework. If it comes from an air-photo or satellite image, its pixels should not be smaller than the resolution of the camera that recorded it.

Data density

Data density is a measure of how many features per area are stored, and may imply a minimum feature size. Greater density implies more features in a given area, and therefore the features may be smaller.

The density of paper map's data is limited by its scale (and therefore its resolution). Areas (polygons) cannot be shown if they are smaller than the lines which draw them. For example, a polygon less than 250 metres wide cannot be drawn on a 1:250,000 scale map. This minimum size also limits the number of polygons that can be represented in a given area of a paper map.

A GIS stores its data digitally, so the minimum size of a feature is limited only by the resolution, which is effectively infinitesimal. Where the degree of detail in a coverage is arbitrary (eg soil polygons), a data definition or convention should specify the minimum size of features, and therefore their density. Without this, different parts of the same coverage may have widely varying degrees of detail, influencing analysis results.

Data detail

Data detail is a measure of how much information is stored for each feature.

A GIS stores lines (eg, a lake shoreline) as a sequence of point locations, and draws it with the edges that join them. There is no limit to how many points can be stored, or how close together they may be.

The amount of detail on line features should be limited just like data density. It does not make sense to store points at intervals which are shorter than the accuracy of their locations.

GIS analysis

In a GIS, analysis is done at the resolution of the data, not at any display scale. For example, the area of a habitat polygon is calculated to the nearest square centimetre. The GIS will carry much more resolution through its calculations than are justified by the data's accuracy. The results of these calculations should be rounded to a value appropriate to the uncertainty of the data for reporting.

Some operations may result in features which are smaller than the data uncertainty. For example, overlaying rivers and forest polygons may create 'slivers' along the riverbanks which are 10 metres wide, when the uncertainty of the data is 20 metres. These slivers should be ignored, or included with their neighbours before the results of the overlay are used for further analysis.

Separation of data and annotation

In a GIS, it is common to display the same data (eg wildlife management unit boundaries) at several different scales for different purposes. It is also possible to create symbols and text that 'look right' at several different scales, and store them apart from the data they label.

For example, management unit boundaries could be stored in one provincial coverage, and annotation layers could be developed for labelling them at display scales of 1:20,000, 1:250,000, and 1:2,000,000.

If done carefully, this avoids duplication of the same data for display at different scales.

Generalization

In a GIS, it is possible to create a new coverage by reducing the amount of detail in existing coverage. This 'generalizing' may or may not reduce the number of objects in the coverage.

For example, a detailed forest cover map may be generalized by combining polygons with similar characteristics. This reduces the number of objects in the coverage.

Conversely, a detailed ecosystem classification map may be generalized by reducing the amount of detail in the boundaries between regions, without reducing the number of regions.

Generalizing a raster image usually reduces both the number of objects, and the amount of detail.

Map series

It is convenient to identify a series of paper maps by their scale (the 1:50,000 water atlas), or the amount of earth they cover (eg NTS 2-degree letter blocks). Neither of these are well-suited to GIS data. GIS data can be displayed at any scale, and can be manipulated as a seamless coverage for either analysis or display.

A GIS coverage should be identified by its accuracy (or uncertainty) and data density (or minimum feature size).