In some cases, the data is not continuous (altitude, slope, distance, etc.), but discrete, i.e., consisting of classes.
Each pixel then corresponds to a category: a soil type, land use, risk level, etc.
These “categorical” rasters are very common in spatial analysis, but they pose a problem:
how can they be integrated into a fuzzy multi-criteria analysis when they do not contain continuous numerical values?
This is where the Classes → Fuzzy module of the FuzzyAttributes plugin (version 2) comes in.
The principle
Rather than working directly with classes (1, 2, 3… or “forest,” “urban,” “water”), we seek to express for each of them the extent to which it satisfies a given criterion.
Examples:
- For a criterion of “areas favorable to infiltration,”
- sandy soils → satisfaction 0.9
- clay soils → satisfaction 0.3
- urban areas → satisfaction 0.1
- For a criterion such as “erosion risk,”
- steep slopes → 0.8
- forests → 0.2
The result is a fuzzy raster, in which each pixel takes a value between 0 and 1 representing the degree of satisfaction of the criterion.
Steps in the FuzzyAttributes plugin
- Menu: Classes → Fuzzy
- Select a categorical raster (single band, integer type).
- The plugin reads the unique values present in the layer.
- For each class, enter the fuzzy membership degree (between 0 and 1).
You can also load a correspondence table from an existing CSV file. - Choose a name for the output fuzzy raster (e.g., fzy_classes.tif).
- Click Create fuzzy raster.
Best practices
- Consistency depends on the criterion studied: the same class may be favorable in one analysis and unfavorable in another.
- Avoid absolute values of 0 and 1 unless you are completely certain: fuzzy logic allows you to express nuances.
- Document your correspondences: they can be saved in a CSV file or a GeoPackage table for reuse.
A concrete example
Objective:
Create a suitability map for installing beehives based on the CORINE Land Cover raster.
Principle:
Each type of land use (forest, grassland, agricultural area, urban area, etc.) is assigned a satisfaction rating between 0 and 1 according to its suitability for beekeeping.
Step 1: Data source
You can download the CLC raster directly here:
https://land.copernicus.eu/pan-european/corine-land-cover/clc2018
Available formats: GeoTIFF, 100 m resolution.
You can then crop it to your study area (e.g., Corsica).
Step 2: Class correspondence
Here is an example correspondence table for bee assessment:
| CLC code | Simplified description | Degree of suitability for bees (μ) | Justification |
|---|---|---|---|
| 111–142 | Urban/artificial areas | 0.1 | Very low, few natural floral resources |
| 211–244 | Agricultural areas (crops, agricultural mosaics) | 0.6 | Presence of meadows and possible honey crops |
| 311–313 | Forests (deciduous, coniferous, mixed | 0.8 | Rich in honey species depending on the season |
| 321–324 | Natural vegetation (scrub, maquis, etc.) | 0.9 | Very favorable, widespread flowering, few disturbances |
| 331–335 | Open spaces, beaches, rocky areas | 0.3 | Few flowers, difficult conditions |
| 411–423 | Wetlands | 0.4 | Unstable environments, sometimes rich but fragile |
| 511–523 | Water bodies/sea | 0.0 | Obviously not favorable |
Step 3: Fuzzy transformation
Using the FuzzyAttributes plugin → Classes module → Fuzzy:
1-Load the CLC raster (reclassified or original).
2-Define the correspondence table above (codes and μ).
3-Generate the fuzzy raster fzy_corine.tif.
4-View: favorable areas (μ close to 1) will appear lighter.
Step 4: Interpretation
- This produces a continuous map of suitability for beekeeping, rather than a patchwork of rigid classes.
- It can then be combined with other fuzzy criteria:
- proximity to water (need for watering)
- altitude/slope
- distance to roads (accessibility)
- absence of treated crops
A fuzzy aggregation of these criteria will produce a multi-criteria map of optimal hive locations.
In summary
Transforming a raster of classes into a fuzzy raster means moving from a “categorical” logic to a ‘gradual’ logic—we move away from “all or nothing” to represent the degree of relevance of each class to an objective.
