GIS Based Studies in the Humanities and Social Sciences - Chpater 19

19 Visualization for Site Assessment Hiroyuki Kohsaka and Tomoko Sekine CONTENTS 19.1 Introduction ...............................................................................................279 19.2 Multilevel Measures of Accessibility and Its Spatial Variation within Residential Districts......................................................................281 19.2.1 Accessibility Measured at the Residential-District Level.......281 19.2.2 Accessibility Measured at 100 M Mesh Level..........................282 19.2.3 Visualization of Spatial Variation in Accessibility within a Residential District.....................................................................284 19.2.3.1 Bivariate Map of Accessibility and Its Variability.....285 19.2.3.2 Composite Map of Accessibility by Two-Level Visualization....................................................................287 19.2.3.3 Accessibility Map at Variable Spatial Level...............289 19.3 Measure of Accessibility by Highly Accurate Simulation and Its Visualization..........................................................................................290 19.3.1 Population as Demand Volume..................................................290 19.3.2 Development of Road Network .................................................291 19.3.3 Measure of Navigation Road Distance by Highly Accurate Simulation Considering Complex Traffic Conditions......................................................................................292 19.4 Conclusion ..................................................................................................296 References.............................................................................................................298 19.1 Introduction Numerous approaches for site assessment have been developed in geogra-phy to evaluate sites for housing and retail facilities (Orford, 1999; Jones and Simmons, 1990). These approaches evaluate a site in terms of two factors, 279 Copyright © 2006 Taylor & Francis Group, LLC 280 GIS-based Studies in the Humanities and Socail Sciences such as the site itself and its location. Site factor is related to the lot in which a facility may be located and the physical environment directly related to the facility. Location factor is connected with its surrounding, which provides opportunity of use or demand. Recently, site-assessment approaches have been performed on GIS to handle very complicated consumer markets (Bir-kin et al., 2004). Accessibility is one of the major elements for the location factor in site assessment. Accessibility is measured from two sides, demand and supply. The measure of accessibility from the residential site to retail and service facilities is related to evaluate a housing site from the demand side. Five types of accessibility measures have been proposed: 1) container index, 2) minimum distance, 3) cumulative opportunity, 4) gravity potential, and 5) space–time (Kwan 1998). Talen and Anselin (1998) point out that the choice of accessibility measure has to be considered very carefully using a case study of the geographic-accessibility measures to public playgrounds at the census-tract level. For site assessment to a retail or service facility, it is necessary to evaluate whether a site will be able to attract a certain volume of sales. Evaluation methods have been developed, such as 1) rating model, 2) regression model, and 3) spatial-interaction model (Birkin et al., 2002). The first is related to compare relative scores for sites, and the second and third can predict the sales volume using mathematical models. Accessibility for supply side is measured in site assessment for retail and service facilities. In the rating model, buffer technique is used to determine a straightforward, “accessible” area followed by overlay technique to clip out this buffer area. However this buffer/overlay approach has some shortcomings, in the point that transport network, natural or man-made barriers, and competition with already estab-lished outlets are not taken into account (Geertman, et al., 2004). However, this approach is widely used in practical site assessment by the reason of its simplicity (for example, site assessment for petrol forecourts is referred to in Birkin et al., 2003). When these site-assessment approaches are applied to practical scenes, many problems have been pointed out. One of the critical issues is the accuracy of analytical results. Inaccurate results cannot be guaranteed to clear the hurdle of a resident’s satisfaction or a client’s sales target. The reliability of the end result to reduce the risk of a wrong or misleading decision is important to site assessment (Van der Wel, et al., 1994). The decision-maker therefore wants to reveal the extent to which uncertainty affects the “decision space.” The presentation of uncertain information is one use of visualization in the GIS community. The extra visual attribute that a visualization environ-ment provides can be used to add a further dimension to a map, in order to judge “truth” on GIS by measures, of uncertainty, error (accuracy), variation, validity, reliability, stability, or probability (MacEachren, 1995). The visual-ization techniques to display uncertainty include side-by-side, overlay, and merged displays (Beard and Buttenfield, 1999). The merged display makes Copyright © 2006 Taylor & Francis Group, LLC Visualization for Site Assessment 281 use of a bivariate map as a representation of quantitative data and reliability of those data. For example, visualization techniques are applied to convey classification of uncertainty in classified imagery and soil maps (Fisher, 1994a; 1994b). For classified imagery, the uncertainty inherent in the assignment of a pixel to a class is conveyed by making the value or color of a pixel proportionate to the strength of it belonging to a particular class. Gahegan (2000) depicts a false color satellite-image fragment of an agricultural area, where vertical offset is used to represent the probability (as determined by a classifier) of a pixel being classified as “wheat.” This paper tackles improving the accuracy of site assessment using suitable visualization techniques to reduce the risk of a misleading site selection. In the second section, the visualization is applied to display classification uncer-tainty in an accessibility map to ophthalmic clinics. The third section per-forms a highly accurate simulation as a site-assessment approach for a car dealer to reveal “truth” as an inaccessible site. The last section discusses a mechanism to judge whether a highly accurate approach should be applied in the practical scenes. 19.2 Multilevel Measures of Accessibility and Its Spatial Variation within Residential Districts 19.2.1 Accessibility Measured at the Residential-District Level As a case study in this section, accessibility is measured from residential districts (Cyocyo-aza) to ophthalmic clinics in Matsudo City, Chiba Prefec-ture, Japan. Matsudo is one of the satellite cities in the Tokyo metropolitan area. Its area is 61 square kilometers, and 19 ophthalmic clinics are located within the city. The shortest-path distance to the nearest clinic is measured using the second method in five accessibility measures mentioned above. Figure 19.1 shows location (+) of clinics and centroids of residential districts (residential point;m) on the road network of the northwest part of Matsudo City. The shortest-path distance from each residential point to the nearest clinic is measured on the actual road network using network analysis of ArcView (Sekine, 2003). Figure 19.2 shows statistical distribution of the shortest-path distance for 343 residential districts. The average of the distance is 1177 m, and its stan-dard deviation is 575 m. By considering such a distribution of the distance, the degree of accessibility is divided into four accessibility levels, as follows: “good” is shorter than 750 m, “normal” is 750 m to 1500 m, “bad” is 1500 m to 2250 m, and “very bad” is longer than 2250 m. All residential districts in Matsudo City are classified into four levels of accessibility in Figure 19.3. Copyright © 2006 Taylor & Francis Group, LLC 282 GIS-based Studies in the Humanities and Socail Sciences 0 1 2 km Clinic Residential Point Road N W E S FIGURE 19.1 Location of ophthalmic clinics and centroids of residential districts on road network. According to this result, residential district “A” shown in Figure 19.3 was assessed as “good” in terms of the accessibility to ophthalmic clinics. 19.2.2 Accessibility Measured at 100 M Mesh Level Now let us measure the accessibility at finer level. The 1-kilometer mesh constructed in the BasicArea Mesh System1 is divided into 10 equal segments for each side to create 100 m mesh. The shortest-path distance to the clinics is measured from the centroids of 6089 100 m meshes constituting Matsudo City. Figure 19.4 shows the four accessibility levels at 100 m mesh level. The residential district A consists of 22 meshes, as shown in Figure 19.5. Fourteen meshes are assessed as “good” accessibility, and eight meshes are assessed as “normal.” Therefore, we can recognize variation in accessibility Copyright © 2006 Taylor & Francis Group, LLC Visualization for Site Assessment 283 3500 3000 2500 2000 1500 1000 500 0 Rank FIGURE 19.2 Shortest-path distance to the nearest ophthalmic clinics from residential points. within this district. An issue may be raised by the residents in the meshes assessed as “normal,” because they will find that their residential place is “normal” in spite of being assessed as “good” at the district level. This is known as modifiable area-unit problem (MAUP) and is particularly impor-tant for the residents in the case of lowering the accessibility level. In this case, the site assessment gives wrong information to them. To examine such a variation in accessibility within all residential districts of Matsudo City, the accessibility map at the district level (Figure 19.3) was intersected with one at 100 m mesh level (Figure 19.4). Table 19.1 shows the variation volume of accessibility between two levels. The diagonal cells represent no change in accessibility level. These ratios are about 70 percent for “normal,” and about 60 percent for “good,” “bad,” and “very bad.” The ratios to the lower accessibility level are 33 percent for “good” and 15 percent for “normal.” Inversely, the ratios to raise the level are 13 percent for “normal,” 28 percent for “bad,” and 36 percent for “very bad.” For good or bad, it became clear that 30 percent to 40 percent of meshes have different accessibility levels from the one measured at the district level. The degradation of accessibility level, in other words, the rate at which spatial analysis at the district level over-assesses, amounts to 15 percent to 33 percent. And more than 30 percent of meshes within the district assessed as “very bad” raise the accessibility level. The result of this analysis means that the accessibility measured at the district level has not enough accuracy in practice. Copyright © 2006 Taylor & Francis Group, LLC ... - tailieumienphi.vn