|
|
The Use of Multi-Attribute Theory and Geographic Information Systems to Support Parcel Ranking and Open Space Prioritization21 July, 1999: View our PowerPoint presentation, Prioritizing and Ranking Land Use. Quantitative Decisions (Merion Station, PA) is helping townships to rank land in for identifying parcels for open space planning. The method is based on techniques of multi-attribute decision theory (Keeney & Raiffa, Decisions With Multiple Objectives, Cambridge U. Press). Ranking involves implicit trade-offs among land valuation attributes such as walking distance to residences, commercial value, ecological attributes, etc. For example, people usually develop ranking systems by assigning numerical scores to each property or parcel of land. Scores are assigned by (a) creating numerical metrics for the attributes, often by placing them into discrete classes, and (b) introducing weights for each class, so that the score for each property is the weighted sum of the numerical attribute values. Ranking is then accomplished simply by sorting the properties by their scores. This approach is good but it lacks several critically important elements. The first is that such scoring is equivalent to making implicit trade-offs among attributes. Those trade-offs should be made explicit and apparent to all stakeholders and decision makers so that they understand what the ranking is doing and can influence it appropriately. The second is that the trade-offs stakeholders WANT to make do not necessarily fit the tradeoffs that such a simple model HAS to make. (This is a big deal: all weighted-sum models make the same trade-offs among attributes regardless of the levels of any other related attributes. Thus you have no mechanism to account for compensating factors.) I begin with the premise that the ranking model should be driven by stakeholder and decision maker needs and criteria rather than limitations inherent in the model formulation. The methods suggested by Keeney and Raiffa allow the decision makers to explore the tradeoffs in a detailed, quantitative fashion (without needing to introduce any decision theory or GIS at all). We can thereby determine (a) what the appropriate metrics should be and (b) how best to make the tradeoffs, all while keeping the model as simple as possible. Often the resulting model is similar to the familiar simple weighted-sum system, in which case what we have accomplished is a rigorous, open, and defensible justification for the model weights. In other cases portions of the model are more complex and are not simply a sum of weighted scores, in which case we have definitely improved on ANY model that the traditional method could produce. GIS (Geographical Information Systems) come in because many of the metrics involve area, distance, amount of coverage, and other geographic measures. For example, Quantitative Decisions recently completed a system to rank over 8,000 properties for the NJ Department of the Treasury. These properties had to be evaluated for suitability to host wireless communication towers. We used the GIS to develop measures of total annual traffic flow within potential coverage areas, total population covered, proximity to ecologically sensitive areas, presence within height-restricted areas, and so on. We also needed some of the more advanced GIS capabilities to implement the ranking model, which involved fairly complex trade-offs between environmental and commercial attributes, conditional on the presence of buildings on each property (the conditionality alone precludes using any simple weighted-sum solution). With a little programming, we found ArcView + Spatial Analyst to be up to the task, which required processing extensive vector coverages of the entire state as well as rasterizing them at about 100-200 m resolution statewide for some of the calculations. (This coarse grid was appropriate since typical tower coverages were assumed to be several square miles. We use statistical techniques to determine things like grid size, scale, and resolution so that the GIS is only as big and complex as it has to be.) The GIS is crucial as a tool to implement an appropriate model, but model development is influenced by the GIS only insofar as having the GIS capabilities gives us the confidence to develop the best models in the first place, rather than resorting to the possibly inferior models people traditionally work with. It is crucial for the model developers to be fully aware of what the GIS can do and how much effort it may take, so they can keep the decision making at a practical level. This is, indeed, an interesting collaboration among GIS analysts and decision makers. Links to related useful sites |
ColorRamp, Memorized Calculations, Rotate, Sample, XSect, and Tissot are trademarks of Quantitative Decisions. All other products mentioned are registered trademarks or trademarks of their respective companies.
Questions or problems regarding this web site should be directed to information
(@quantdec.com).
|
