What is suitability analysis?
Site selection questions have the same shape. Where could a new clinic, warehouse or school go, given that it should be near roads, away from existing facilities and off steep ground? Suitability analysis ranks every location by how well it meets a set of criteria. It is a form of multi-criteria analysis, and in GIS it is usually done with rasters.
The output is a map of scores. It shortlists areas for a closer look. It does not make the decision, because ownership, price, zoning and local knowledge are not in the data.
Distance surfaces
Many criteria are distances, such as the distance to the nearest road.
A Euclidean distance raster stores, for every cell, the straight-line distance to the nearest source feature. It is fast and needs only the source features.
A cost distance raster stores the accumulated cost of travelling to the nearest source, summed over the cells crossed. It needs a cost raster that says how hard each cell is to cross. A flat cell might cost 1, a steep cell 5, and a lake cell has no value at all, which makes it a barrier. Cost distance suits questions about effort, such as building a road or walking, where straight lines ignore the terrain.
Steps of the analysis
- Define the goal and list the criteria. Each criterion must be something you can map.
- Prepare one raster per criterion on the same grid.
- Standardise each raster to a common scale, such as 0 for worst and 1 for best.
- Choose a weight for each criterion, with the weights adding up to 1.
- Multiply and add the scores. The result is the suitability raster.
- Apply constraints, then pick the highest scoring cells.
- Test how the answer changes with different weights.
Standardising
A distance in metres and a slope in degrees cannot be added. A simple rule stretches each one onto 0 to 1. Slope can score 1 on flat ground and 0 at 8 degrees or more. Road access can score 1 on a road and 0 at 1.5 kilometres or more. Some criteria run the other way. If the aim is to serve places that existing hospitals miss, the distance from existing hospitals scores higher the farther it is. Check the direction of every criterion.
Weights and constraints
Weights express how much each criterion matters. They are judgements. A common way to set them is to compare criteria in pairs, as in the analytic hierarchy process. Whatever method you use, write down the reasons.
Constraints are different from criteria. A criterion makes a cell better or worse. A constraint removes a cell altogether. Lakes and protected parks are usually constraints, since no score makes them available. Apply constraints as a mask, setting excluded cells to no value.
Sensitivity
Run the analysis again with other weights and see whether the best areas stay the same. If a small change in weights moves the best sites a long way, the ranking is fragile, and you should say so.
Example: a gap in hospital coverage
The QGIS tutorial scores three criteria for a 30 by 20 kilometre window of Bengaluru: slope, distance to the nearest road, and distance from existing hospitals. The weights are 0.2 for slope, 0.4 for road access and 0.4 for the hospital gap. Lakes and parks are masked out. Scores run from about 0.21 to 0.99, and the top 5 percent start at about 0.74. The best cells sit in flat areas that are near a road and far from an existing hospital. They are candidates for a visit, not a conclusion.
Common mistakes
- Adding factors without putting them on a common scale first.
- Getting the direction of a criterion wrong.
- Counting two correlated criteria separately, so one idea gets double weight.
- Treating weights as facts instead of choices.
- Forgetting constraints, so a lake scores as a good site.
- Measuring distance on a window that cuts off features just outside it.
- Reading a ranking as a recommendation.
Glossary
| Term | Meaning |
|---|---|
| Criterion | A factor that makes a location better or worse |
| Constraint | A rule that excludes a location completely |
| Standardisation | Rescaling criteria to a common range such as 0 to 1 |
| Weighted overlay | Multiplying criteria by weights and adding them |
| Cost surface | A raster of how hard each cell is to cross |
| Sensitivity analysis | Testing how the result changes when the inputs change |
Practise it
The tutorial Distance surfaces and site suitability builds each of these layers from OpenStreetMap and a DEM.
Test yourself
1. What is the difference between a criterion and a constraint?
A criterion changes how good a location is. A constraint removes the location from consideration.
2. Why does cost distance differ from Euclidean distance?
Cost distance adds up the effort of crossing each cell, so it bends around barriers and stretches over steep ground. Euclidean distance is a straight line.
3. Why test different weights?
Weights are judgements. If the best sites change a lot when the weights change a little, the ranking is not robust.