Raster vs Vector GIS – Introduction
|Map Accuracy||Pedigree of original paper map needs to be considered (ie. the scale and quality of the source map).||Pedigree of original paper map needs to be considered (ie. the scale and quality of the source map).|
|Map Accuracy||Grid cell size a key determinate of accuracy||Supports surveyor level accuracy|
Digital Elevation models
Google map themes (roads, addresses, etc)
Census and social sciences
|Database||Each individual cell is represented as a value||Each geographical object has at least one value attached to it in a database. The database might be a simple flat structure, or it may have a very sophisticated design.|
Raster Geographical Information Systems and vector Geographical Information Systems operate differently and are used to address different types of geographical problems. The Raster vs Vector GIS question is rarely one you need to worry about these days. The geographical problem you’re trying to solve and the maps you have to solve the problem most often make that decision for you. Much of the angst about Raster vs Vector GIS needs to be seen in the context of the flakey GISs, flakey computers, inefficient computer algorithms and slow computers that existed in the early days of GIS. These concerns were real at the time, but no longer apply. For example, Digital Elevation Models models (DEMs) that took hours to produce in the early days of GIS, now complete in seconds.
Broadly speaking, vector GISs are most suited to applications that involve the social and engineering sciences, and grid cell (raster) systems to applications that involve the natural sciences. Vector GISs are used to display maps of roads, land ownership and buried infrastructure – the sorts of maps that you’re used to seeing in google maps.
Most vector Geographical Information Systems can display bitmap backdrops of air photographs and satellite images. However, the primary purpose of the raster GISs is to analyse the values of each pixel rather than just display them.
Raster GISs are also known as grid cell GISs and sometimes pixel GISs. They are used to represent geographies that vary continuously – things like digital terrain models, subsurface water models, weather models and satellite imagery (a discussion of models is a whole other topic).
A vector GIS works by storing three types of geography – points (a power pole), lines (a power line) and polygons (a power utility service area) in a geographical database, and their attributes are stored in a separate database
Vector GISs have databases attached to them. Some databases are simple and some are very complex. At its simplest, the database functionality allows users to do things like click on a land parcel and find out who owns it.
Professionally surveyed vector GIS maps can be precisely accurate, but vector maps that have their origins in paper maps will reflect the pedigree of the original map. By pedigree, I mean was the GIS map digitized from a paper map or a stable based map, was the original map large scale or small scale, what was the purpose of the map, etc.
Raster (Grid Cell) GIS
Grid cell (raster) Geographical Information Systems work by storing attribute data as grid cell values. Grid cells are an ideal data structure for applications involving terrain analysis. They have superior analytical power to vector GIS, but grid cell map presentation tends to be less attractive than vector map presentation.
The grid cell model is ideal for applications where dynamic buffering (such as a distance from a roadway or watercourse), the inclusion of remotely sensed information, or the use of spatial modelling techniques is required.
Appropriately used, grid cell GIS overcomes the spatial variation problem that the vector GISs have difficulty coping with (figure 1). For example, traditional slope maps show slopes as polygons (eg. >10% slope and < 15% slope). Although you can tell that a polygon contains slopes within the given range, it is impossible to tell the exact slope at any point in the polygon. In a grid cell GIS, each cell in a slope map derived from a DEM has an individual value at the resolution of the grid cell.
Selection of Cell Resolution
Grid cell maps suffer from the same pedigree issues as vector maps, but in addition they suffer from grid cell size related issues.
Raster GISs grid cells can be any size. For example the cells in the Lidar terrain map might be one metre. Pixels in Landsat imagery can be as large as 60 metres and in other satellite imagery, pixels might be two or more kilometres.
Grid cells are square, and, unless a study area is either square or rectangular, an image will contain redundant grid cells (figure 2). To ensure that statistical analysis is confined to the study area, null values are assigned to the grid cells outside the study area.
The polygon in figure 2 is irregularly shaped and the dimensions of the grid cell map is determined by the extremities of the study area (the minimum and maximum X and Y coordinates). In this example the extremities of the study area mean that the grid cell map covering the field area of interest must represent an area 100km by 150km.
Cell resolution refers to the size of a grid cell and is defined as the length of an axis / number of cells along the axis. So a 100km (10 cell) * 150km (15 cell) grid cell map would have a cell resolution of 10 kilometres (100km2) (figure 2).
Because grid cell systems represent terrain features at the resolution of the cell, the identification of a suitable cell resolution is a major consideration in implementing a grid cell GIS, and will undoubtedly be one of the most difficult tasks undertaken by the GIS analyst. The cell resolution you choose will most likely be situation-dependent, but you will also need to consider the original scale of the base data and the issues that the data needs to address. You can overcome many potential problems through careful analysis.
The raster v’s vector GIS conundrum is not a conundrum at all these days. The geographical problem in combination with the GIS maps that you have available to address the problem most times will make the choice an obvious one.
Vector systems are usually better suited to engineering, government and social science applications. Grid cell systems are usually better suited to environmental applications.
The way data are stored in the two systems is quite different. In a grid cell GIS the data are embodied in the geographical file. In a vector GIS the data are stored in separate data tables.
The quality of the spatial data are a big difference between the two. Vector GISs are capable of survey level accuracy, while grid GISs accuracy is determined by the resolution of the grid cells. In addition, both raster GIS and vector GIS are hampered by the pedigree of any paper map that finds its way into the system.