A suitability model is a model that weights locations relative to each other based on given criteria. Suitability models might aid in finding a favorable location for a new facility, road, or habitat for a species of bird. Overlay analysis is a common method for creating a suitability model which involves using GIS techniques and software. Overlay techniques were originally advanced by Ian McHarg who used a manual overlay cartographic process which he describes in his 1969 book Design with Nature. With the advancement of computer mapping software, suitability modeling has become much easier and faster to implement, and today it is used for many varying tasks.
There are seven general steps required to create an acceptable suitability model:
- Define the problem
- Break the problem into submodels
- Determine significant layers
- Reclassify or transform the data within a layer
- Weight the input layers
- Add or combine the layers
Without a clear understanding of the problem that needs to be solved a suitability model cannot be successful. All other steps in the process will contribute to the objective of solving this problem. The components of this objective should also be defined, as well as a way of knowing when the problem has been solved. Consider the issue of deforestation, to lower deforestation rates a suitability model could be created to model areas most likely to be deforested in the immediate future; laws and regulating entities could then be focused on those areas most susceptible to deforestation. The overall goal of the deforestation suitability model would be to slow the rate of deforestation.
The complexity of most suitability modeling problems can be overwhelming and confusing; for this reason, it is advisable to break the model into submodels. For deforestation there are many different drivers, therefore a variety of submodels would be needed. Population, population density, movement of people, elevation, slope, land cover type, hydrology, location of protected areas, soil type, laws, roads and infrastructure, the list could go on, all of these things affect where deforestation happens and the intensity. Combining these factors could lead to a submodel for physical environment (elevation, slope, land cover, land use, soil type, and hydrology), for built environment (roads, infrastructure, and other relevant transportation networks), and for demographic characteristics (population, population density, population growth rate, and poverty rate) .
Each submodel should be defining an aspect of the overall model, and only submodel factors which contribute to solving the original problem should be included in a submodel. It is in this step that data must be gathered and layers created; for example, it may be known that deforestation usually happens a certain distance from city/road/agricultural areas, therefore a Euclidean distance tool (within a GIS software package) could be used to create a distance raster around these areas.
There are many different datasets going into the model, all with varying number systems; this means that attempting to combine these datasets would give meaningless results. Therefore, a common number scale should be chosen (usually 1 to 9 for a weighted overlay and 0 to 1 for a fuzzy overlay; with larger values signifying more favorable areas) and each dataset reclassified to the new scale (there should be a tool for this in most GIS applications).
If there is strong evidence that some factors contribute more to the main goal these factors should be weighted based on their level of contribution. For instance, focusing specifically on deforestation in Africa, previous research shows that one of the main causes of deforestation is fuel wood extraction; therefore, variables associated with fuel wood extraction should be weighted more heavily than other variables. It should be noted that weighting should not be done if a fuzzy overlay is used.
To complete the model, all factors must be combined, usually through a weighted overlay or fuzzy overlay technique. For a weighted overlay all the factors would be added together and reclassified to form a new data layer where high values signify more favorable locations and low values less favorable locations. A fuzzy overlay analysis produces the same type of results but through more complex methods.
Once the suitability model is complete the results should be analyzed. It is always a good idea to examine the results closely to verify that they make sense and no mistakes were made. Before the model is used the results should also be verified and validated. After the analysis is complete locations can be selected using the model and this information can be applied to the original problem.