Raster Multi-Criteria Analysis Tools

This is intended to be a set of tools for performing linear weighted multi- criteria analysis on rasters describing different features of a given geographical area.

Technical requirements

1. Currently requires ArcGIS 10.0 and arcpy with a valid license for the Spatial Analyst extensions. Sadly this means the script will only run on Windows. A future version might be able to remove this dependency on ArcGIS.

2. arcpy is available in your PYTHONPATH. To verify, open a python shell and type the following command:

import arcpy

If this executes without error then you're in the clear. If not, ensure that PATH and PYTHONPATH are set correctly. This ESRI help page might be of assistance:

http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//002z00000008000000.htm

combine_fuzzies.py

This script assumes you have a number of fuzzified raster files covering the same geographic extent, with the same spatial reference system and projection. These should have been normalised to a scale with values between 0 and 1, based on whatever fuzzy membership function you choose (most likely a monotonic, trapezoidal, triangular or gaussian function). At present we do not provide a script for generating these fuzzified rasters, although that would have been awesome.

The combine_fuzzies.py script itself enables a linear weighted multi-criteria analysis to be performed on the resulting fuzzified raster files.

To use it:

1. Copy raster files into position

In this version the raster files must be in the same directory as the combine_fuzzies.py script (possibly not a very good assumption).

2. Configure source raster files

Open source_files.py. This defines a list called SOURCE_FILES. Each element in the list is a dictionary representing one fuzzified raster file, and having the following members (i.e. key/value pairs):

• name: an arbitrary name
• file_name: name of the raster file on your file system
• overlay: name of the fuzzy overlay method to be used. Currently supports 'SUM' and 'AND'. Possibly others too, but they are untested.
• invert: a flag indicating whether the fuzzified raster should be inverted prior to processing. This is useful in cases where ideal values are 0 rather than 1.

For example:

SOURCE_FILES = [
    {'name': 'fauna', 'file_name': 'fuz_fauna2', 'overlay': 'SUM', 'invert': False},
    {'name': 'flora', 'file_name': 'fuz_flora2', 'overlay': 'SUM', 'invert': False},
    {'name': 'slope', 'file_name': 'fuz_slopeper', 'overlay': 'SUM', 'invert': False},
    {'name': 'fire', 'file_name': 'fuz_fire1', 'overlay': 'SUM', 'invert': True},
    {'name': 'ftrail', 'file_name': 'fuz_firetrail', 'overlay': 'SUM', 'invert': False},
    {'name': 'erosion', 'file_name': 'fuz_erosion1', 'overlay': 'SUM', 'invert': True},
    {'name': 'mangroves', 'file_name': 'fuz_mangr2', 'overlay': 'AND', 'invert': False},
    {'name': 'creeks', 'file_name': 'fuz_creek4', 'overlay': 'AND', 'invert': False},
]

3. Configure scenarios

A scenario defines a set of subjective preferences for the values described in each raster file. These preferences are encapsulated in a list called SCENARIOS. This list contains one or more dictionaries, each of which must contain this member:

• name: an arbitrary name for the scenario

After that, each dictionary in SCENARIOS must have one member for each of the source rasters described in SOURCE_FILES. These data members are key/value pairs and must be of the following form:

• key: must match the name of a source file defined in SOURCE_FILES
• value: a weighting value between 0 and 1

Ideally all weighting values in a scenario should add up to 1, although this is not enforced in the script.

For example:

SCENARIOS = [{
    'name': 'conserv',
    'fauna': 0.3,
    'flora': 0.3,
    'mangroves': 0.3,
    'creeks': 0.02,
    'slope': 0,
    'fire': 0.04,
    'ftrail': 0.02,
    'erosion': 0.02
}, {
    'name': 'min_cost',
    'fauna': 0.01,
    'flora': 0.01,
    'mangroves': 0.03,
    'creeks': 0.1,
    'slope': 0.3,
    'fire': 0.1,
    'ftrail': 0.25,
    'erosion': 0.2
}]

4. Execute script

To start the process just run the following command:

PATH_TO_YOUR_ESRIFIED_PYTHON\python.exe combine_fuzzies.py

Or on systems with PATH and PYTHONPATH correctly configured, you can instead just double-click on the combine_fuzzies.py file from Windows Explorer.

One output raster will be calculated for each scenario, based on the preferences you defined. They will be found in the output/ directory (which is a subfolder of the directory containing combine_fuzzies.py).

5. Re-running the script

combine_fuzzies.py does not attempt to re-generate output rasters if an output file for a given scenario already exists. If you want to rerun the script with different values you must first delete or move the relevant output file(s).

TODO

• Make input and output paths configurable, or pass them in via command line options.
• Properly handle fuzzy overlay methods other than 'SUM'. At present the script only works when 'AND' sources files are listed after the 'SUM' files.
• Optimise the fuzzy overlay step. sa.FuzzyOverlay() allows an arbitrary number of rasters to be passed in at once. We should make use of that.
• Write a script to generate (fuzzify) the source rasters.
• Handle different geographic extents, spatial references and/or projections.
• Automatically identify positions or areas with the highest overall score across multiple scenarios.
• Need to write some unit tests. Of course.
• Rework this to use open source tools rather than relying on proprietary ESRI software. Ideally before my student evaluation of ArcGIS expires.