SUITABILITY ANALYSIS

How can we model the area suitability for a particular use?

• Consolidation of critical data tied together by their location using GIS.
• Multi-criteria decision analysis (Weighted overlay) => Agent-based modeling, CA etc.
• relevant criteria + weight correctness = analysis credibility

Carl Steinitz (1995) described the planning processes through 6 six models (see figure below). First three assess the status quo, the second three prupose the intervention to the landscape.

The suitability analysis predicts a scenario for intervention. Firstly, however, a thorough investigation of the status quo is essential to bring relevant outcomes. Subsequently, possible implications for change, imapct, decision models must be considered before running the analysis, and evaluated from the suitability map.

Steinitz (1995) claims that "not all geography changes by design, and not all design changes geography." In the light of the knowledge it is vital to reiterate and test the models to make most educated decisions for sustainable design... that is how to Change Geography by Design.

Site selection probem vs. Site search problem (Cova & Church, 2000) set of potential sites vs. study area ranking the sites vs. setting the boundaries

* study area of Ukarumpa, PNG * settlement expansion model (site search problem) - I.e. search for 3 most suitable sites; 10 ha area each * (time frame) * (to evaluate the impact of the change)

• Complex representation of a location broken down into simple models.

• Bringing relevant and sustainable results requires utilizing wide range of models.
  • natural factors (topography, conservation), human factors (development potential and risk), soft factors (values, needs)

1. What information relates to our question?
2. What models can become a partial answer for the question.
3. What data is at hand?

Data preprocessing in this case means the data rerieval to raster format. The largest cell size of the criteria layers determes the final resolution!

• Distance, density (vector to raster transformation):
  • Kernel density estimation, Euclidean Distance.
• Calculation on raster (derivations, map algebra):
  • slope, terrain roughness, geomophology, hydrology, viewshed.
• Reclassify (extraction of new data):
  • binary: masking with sharp deliniation,
  • fuzzy: expressing uncertainlity, non-sharp boundaries.

Criteria can be translated into the overlay analysis only if are standardized into one common scale.

Suitability Modeler (Spatial Analyst ArcGIS Pro)

• setting the scale to "1–10" (or other);
• histogram clarifies the distribution of the dataset, underlay information of transformation selection;
• transformation definition:
  • unique categories: reclassify classes to the scale, disabled exclusion of a class from the analysis, e.g. land cover data,
  • range of classes: simplify range of values into classes and ranking them, e.g. slope,
  • continuous functions: reclassify based on the function development with function adjustments (linear, logaritmic, exponential,...)

The following gallery of images explains the parametrization for the data standardization used in the case study; contains the preview in the map window, histogram, transormation settings and the distribution of the reclassfied values in the suitability scale (1–10).

• by pecentage: shares have to make it together 100 %,
• by multiplier: no sum requirement.

Wrong weights => wrong results => wrong decisions. It is the EXPERT'S RESPONSIBILITY. => To provide a detailed specification of the analysis workflow.

The case datasets include the natural and human factors. Two conceptual groups of datasets were made: landscape characteristics, accessibility to services. Therefore, each group makes together equal sum of five, wight of ten in total (see figure).

Regarding the landscape characteristics, the land cover is wighted slightly more than the slope. The land cover preconditions are expected to be more determining for the settlement expansion than the terrain jaggedness.

The accessiblity of servicies are divided into the settlement-related (density of buildings and distance to newly built-up areas) and potential-related (proximity to roads and water streams). The settlement-related data is considered more influential than the potential-related data. Therefore, the data is multiplied by the weights three and two, divided equally (see figure).

• manual methods: shades of grey or colors, the large dataset limits
  • Design by Nature (McHarg, 1969)
• SYMAP and GRID systems: Harvard Laboratory
  • flood-control reservoir and parkway for their suitability for recreation and other land uses (Murray et al., 1971)

• Cartograhic modeling and Map algebra (Tomlin, 1990)
  • Boolean overlay, Weighted linear combination (WLC)
• Fuzzy overlay (Burrough & McDonnell, 1998)

The tool identifies the best regions from an input suitability (utility) raster that satisfy a specified evaluation criterion and that meet identified shape, size, number, and interregion distance constraints (Esri, 2022b).

In the question specification the case study searches for 3 possible areas of 10 ha (30 ha in total). In the tool one can specify the degree of freedom for the requested areas (setting the region maximum and minimum area).

The shape-utility tradeoff identifies the weight for the cells when growing the candidate regions in the parameterized region-growing algorithm (PRG). The weighting is a tradeoff between a cell's contribution for maintaining the circular shape relative to the utility contribution of the cell's value (Esri, 2022). The results of 30 % and 70 % tradeoff are compared in the results.

The evaluation method specifies the rule by which the output results are localized. For the comparison the study examines three of them: - the highest sum of the cell values, - the highest median value of the cell set, - the greatest edge maintaining the circular shape.

The region growth is determined by the existing locations (built area mask) and brings the locations preserving the pre-set parameters. PRG calculation starts at the seed cells and grows preserving the given shape and maxizes the utility. The combinatorial selection method tests all the combinations of the number of regions from the candidate locations and selects the best suitable combination. An alternative to the methos is the sequential selection-

The results presented in three swipe-enabled maps. The user can swipe between two diffrent basemaps: the suitability raster and the satellite imagrey.

In each map one can compare the diffrent parametrization of the shape-utility tradeoff having the same evaluation method. The three suitalbility areas have been converted from raster to vector format to calculate their area (see in pop-up window).

None of persented results is better or worse. The first two evaluation methods bring similar results. Although, it does not imply that the third approach is wrong, it is only about much different method of evaluation. Here enters the important role of public participation. The discussion with local planners and inhabitants is inevitable. This will initialize the iterative flow for redesgning the models as Steinitz suggested in his framework (1995).

So far the soft factors dominated the development decision-making of the stakeholders resulting in sprawl effects. However, such rigorous suitabilty analysis can create unbiased base for discussion for sustainable settlement and community development.

The suitability results are highly dependent on parametrization of the analysis. Therefore, the expert's subjective inputs, e.g. wighting, citeria selection, have to be justified to depth by the expert. Every expert has different topical emphasis, so the mutual discussion can bring more clarity and objectivity to the result.

The resolution of the results should follow the intent of the main question. The finer resolution does not imply better results. Sometimes the generalized results keep the "big picture" focus and foster less distraction in the further expert discussion.

Malczewski (2004) discusses the notion of "Collective Design" on how the spatial science evolves from closed positivistic reasonings to open communicative resolutions. "Any planning process must focus on a mix of hard and soft information. Central to the land-use suitability analysis is the way in which these two types of information are combined as well as the right balance between the amount of hard and soft information used in the analysis."