(4) Spatial Analysis

What you can find out in this section:

• Pairwise Comparison Matrix for model 1 & model 2
• Single Objective Suitability Maps
• Cartographic Models for the Suitability Maps
• Post-Aggregation Constraints
• Cartographic Models for the Post-Aggregation Constraint Maps
• Solutions for Multi-Objective Land Allocation Problem
• Cartographic Model for the final MOLA image
• Policy Implications

SCENARIO 1: Maximum Environmental Protection Model

• Rating of Row Factor Relative to the Column Factor

• The Corresponding Factor Weights that are Derived from the WEIGHT Module

• Rating of Row Factor Relative to Column Factor

•    The Corresponding Factor Weights that are Derived from the WEIGHT Module

    After the factor weights are generated and the consistency ratio is within the acceptable range (i.e. less than 0.10), I can input these factor weights together with the constraints into the Weighted Linear Combination (WLC) to undertake the Multi-Criteria Evaluation (MCE).

    The results of WLC are 2 suitability maps -- one for each scenario. These maps indicate the relative suitability (0 = least suitable & 255 = most suitable) of the areas in GVRD for future developments. Notice that a lot of the areas are masked out (i.e. have a value of 0) because of the constraints that are applied.

    For the "Compact Cities" scenario, I did one suitability map without the population constraint (total 3 constraints) and one with the population constraint (total 4 constraints) so that I can notice the impacts of adding in the population constraint in the analysis.

• Suitability Map for Scenario 1

 

• Suitability Map for Scenario 2 -- without the population constraint

 

• Suitability Map for Scenario 2 -- with the additional population constraint

    There are two basic methods for site selection using a continuous image of suitability: Specifying a suitability threshold and total area threshold. Applying either type of thresholds will result in a Boolean map indicating the selected sites.

• Suitability Threshold for Scenario 1
• Goal: To identify areas with a suitability score of 200 or higher.
• I run RECLASS with the suitability map for scenario 1 as the input image -- areas with a suitability score of 200 or above get a new value of 1 and all other areas get a new value of 0.

•  I try to use GROUP with the reclassed image but there are just too many groups being created (even with diagonals) that it would not display the results; however, this outcome with GROUP implies that the solutions (or land allocations) are not spatially contiguous.

• Total Area Threshold for Scenario 2
• Goal: To identify the best 3000 hectares for development under the Compact Cities model.
 
• STEP 1: All pixels are ranked by their degree of suitability.

I run RANK with the suitability map for scenario 2 as the input image to uniquely rank all locations in terms of suitability from high to low (descending order).
•  The metadata indicates that the resolution of the GVRD base map is 50m. Each pixel has an area of 2500 m²; therefore, a hectare (10000 m²) is equal to 4 pixels.
•  In order to have an area of 3000 hectares, the number of pixels required = 3000 * 4 = 12000.
 
• STEP 2:  Pixels are selected based on their suitability.

I run RECLASS with the ranked image such that those cells ranging from 1 to just less than 12001 have a value of 1 and those cells ranging from 12001 to 9999999 have a value of 0.
 
• STEP 3: Creating an image that shows the selected areas in terms of suitability scores.

Finally, I OVERLAY multiply the reclassed image with the final suitability map of scenario 2 to obtain the following image:

 

• As this map shows, the areas that satisfy the total area threshold tend to have a high level of suitability as well.

 

Can the Maximium Environmental model coexist with the Compact Cities model to result in mutually benefitial land allocations? Read on to see the capability of MOLA to resolve this multi-objective problem!

 

Is there any conflict between the areas selected under the 2 scenarios? In other words, would one pixel be the best allocation for both of the scenarios at the same time?

I apply CROSSTAB to the 2 ranked images and obtain the following result:

It may be quite hard to see the red-coloured pixels in the previous image (located in Surrey). Here I zoomed into the areas of conflict to get a better view:

I have repeated the cross-classification again, but this time I CROSSTAB the ranked images using the suitability map of scenario 2 with the population constraint added.
Interestingly, there will be NO CONFLICT at all!

Next, I run the MOLA module to find out a compromise solution that is best for the overall situation -- I specify an equal weight of 0.5 for the 2 objectives, an area requirement of 6000 pixels for the 1^st objective (1500 hectares) and 12000 pixels for the 2^nd objective (3000 hectares), and a 100 pixel tolerance.

• This final image uses the final suitability map of scenario 2 (Compact Cities scenario) without the population constraint. MOLA completes the query in one pass, achieving 5957 and 11993 pixels for each objective respectively.

Both MOLA maps identify the solutions for both of the objectives:

(1) The best 1500 hectares for "Maximum Environmental Protection" development

(2) The best 3000 hectares for "Compact Cities" development

 

The images presented above show the spatial patterns of land allocations under each scenario. However, it may be a little difficult to find out the answers to the following questions visually,