Using GIS to Model and Visualize Congestion Effects on Individual Accessibility

Joseph Weber, Department of Geography, Ohio State University

---

Abstract
    Considerable attention has been devoted to the measurement of accessibility to employment, shopping, educational opportunities, health care facilities, and other services within cities. The use of Geographic Information Systems has enormous utility for such research because of its ability to not only represent the components of the urban environment, such as the home locations of individuals, employment opportunities and retail or other service locations, but also for modeling the spatial relationships among these components through the use of computationally intensive transport network analysis methods. The value of Geographic Information Systems is especially apparent with the use of disaggregate space-time accessibility measures because of their requirement for a very high degree of temporal and spatial resolution of the urban environment, and especially of the accurate representation of the movement possibilities of individuals through urban networks. While considerable attention has been directed at the representation of the urban environment it is argued here that accessibility research has not yet taken full advantage of the network analytical capabilities available within Geographic Information Systems. Instead, even when detailed representations of networks are used, potentially unrealistic measures of travel time based on assumptions about constant travel speeds through the network may be incorporated within studies. It can be argued that doing so creates limitations for accessibility measures as utilizing a single travel time for all hours of the day does not allow for the existence of daily congestion or hourly variations in traffic volumes. Applying a constant travel time to all areas of a city also does not allow for highly localized congestion within transport networks so that traffic flows and the effects of peak hour congestion are uniform throughout the entire urban area. The ability to incorporate spatially and temporally specific traffic congestion is therefore likely to offer considerable insight and detail into individual accessibility. This research seeks to show how these limitations can be overcome by measuring accessibility using space-time concepts with a detailed street network for the Portland, Oregon, metropolitan area, using spatially and temporally varying estimates of highway travel times. Further, because the measurement of accessibility is based on actual travel diary with trip data for 200 individuals, it is possible to incorporate the locations and times of day during which travel took place for each individual. The resulting accessibility values therefore reflect not only each individual's daily activity patterns and constraints, the opportunities available to them in different locations of the city, but also the uneven spatial and temporal effects of congestion. These effects can be visualized by the use of network potential path areas to show the areas and potential activity opportunities which individuals would be able to reach during their travel, both with and without congestion effects. The use of standard ArcView GIS is fundamental to this application because of its network analytical abilities and the need to incorporate the spatial relationships existing between streets, activity locations, and activity opportunities contained in multiple data sets. 

---

Introduction

    Geographers have devoted considerable attention to the study of individual accessibility, whose accurate measurement depends on the accurate representations of the urban environment and possibilities for movement in this environment. The use of Geographic Information Systems (GIS) has enormous utility for such research because of its ability to represent the components of the urban environment (such as the home locations of individuals and employment opportunities) and to model the spatial relationships among these components through the use of network-based geocomputational methods. This is especially so with the use of space-time accessibility measures, due to their requirements for a high degree of spatial and temporal resolution of the urban environment, and their need to accurately represent the movement possibilities of individuals through urban transport networks.

    In this paper, we show that previous studies on accessibility suffer from several limitations that can be overcome through utilizing the representational and geocomputational capabilities of GIS.  While previous work has assumed constant travel speeds throughout the day, this ignores the daily variations in travel speed due to congestion. Also, applying a constant travel speed to all areas of a city is problematic because it assumes that the effects of peak hour congestion are uniform throughout the entire urban area and affect all people equally. Further, past studies also ignore the effects of the business hours of urban opportunities by assuming them to be available throughout the day.  The capabilities of GIS to represent localized traffic congestion and the opening hours of urban opportunities therefore can offer significant insight into the ways that the accessibility of individuals and social groups are affected by not only geographical location but also temporal variations in travel speed and facility opening hours in an urban area.

    The purpose of this study is to show that the incorporation of locally specific travel times within a street network does in fact allow a significant increase in the ability to realistically evaluate individual accessibility within cities.  Using an activity-travel diary data set this research shows that individual accessibility within Portland is not homogenous, and neither does access to employment or shopping opportunities vary according to common monocentric and polycentric expectations about urban form and human behavior.  Instead, the role of distance in ordering or predicting accessibility variations within cities appears to be quite limited relative to variations in individual travel behavior, mobility offered by the street network, and variation in the locations and size of activity opportunities. This paper also shows that incorporating time into accessibility measures in the form of evening congestion and business hours leads to additional (and highly spatially uneven) reductions in accessibility, revealing that time is very important to accurately assessing individual accessibility, and perhaps as important as space.

Data and Procedures

    This study uses a range of data sources. These include an activity-travel diary data set of Portland, Oregon, a digital network model with estimates of free flow and congested travel times, and a comprehensive geographic database of the study area. The analytical procedures involved creating a realistic representation of the temporal attributes of the transport network and urban opportunities in the study area, as well as developing a geocomputational algorithm for implementing space-time accessibility measures within ArcView GIS.

    To measure individual accessibility, data for both individual activity-travel behavior and the location and business hours of activity opportunities in the Portland metropolitan area are required. Individual activity-travel data for the Portland metropolitan area was obtained from the Household Activity and Travel Behavior Survey carried out during 1994 and 1995 by the Portland metropolitan government (Cambridge Systematics 1996). This is a highly detailed two-day travel diary survey that recorded all activities (and activity locations) involving travel and all in-home activities with a duration of at least 30 minutes for over 10,000 individuals in the sampled households.  However, due to the computational intensity of the GIS algorithm used to compute space-time accessibility, only 200 individuals who traveled exclusively by the automobile during weekdays from the original sample were selected for this study (Figure 1).  The sample includes 101 males and 99 females from 187 households. Of these, 157 are employed full time, 28 part time, and 15 are not employed or retired. The sample is racially homogenous, as almost all individuals (185) are European Americans (white).

Figure 1: Portland, Oregon, Study Area

    To represent potential activity opportunities in the study area, a geographic database containing almost 28,000 centroids of commercial and industrial land parcels in the Portland metropolitan region was assembled from local land use data. For computing accessibility measures, attractiveness of individual opportunities in the study area was represented in two ways. One is the area of each land parcel (in acres) that takes into consideration that some activity opportunities are considerably larger in size and therefore more attractive than others. In addition, because buildings located in downtown and major suburban centers often have multiple floors and higher ratios of building size to parcel size, the square footage of these parcels was weighted to take this into account. This weighted area was the second measure of attractiveness used in the accessibility computation. Figure 2 shows the opportunity density surface of the study area using the weighted area of each opportunity.

Figure 2: Weighted opportunity density surface of Portland study area

    The digital street network used in this study is an enhanced U.S. Census TIGER street network that covers the four counties of the study area (i.e. Clark, Clackamas, Multnomah and Washington). While the TIGER functional classification for each link could be used to estimate free flow speeds and capacity, this would still leave peak traffic volumes unspecified.  A variety of methods have been used for estimating intraurban travel times through street networks, including the use of interzonal or centroid to centroid travel times (Muraco, 1972; Wachs and Kumagai, 1973; Black and Conroy, 1977; Knox, 1978; Handy, 1993; Geertman and Van Eck, 1995; Scott, 1998; Helling, 1998; Wang, 2000).  Others have measured direct point to point travel times using assumed speeds for various modes of travel (Lenntorp, 1976, 1978; Miller, 1982), and also sometimes taking into account assumptions about driving speeds over different street types (Brainard et al 1997, 1999; Kwan, 1998, 1999a; Kwan and Hong, 1998).  However, these methods either lack the precision necessary for space-time measures, or do not allow for the effects of congestion to be incorporated.  To estimate link-specific travel speeds under both normal and congested conditions at different times of the day, additional data from a planning network used by the Portland metropolitan government (Metro) for transportation modeling was therefore used. Although this planning network contains only the major streets and freeways and lacks the spatial and temporal resolution needed for space-time accessibility measures, it was useful for estimating free flow and peak period link-specific travel times.

    Free flow travel times were taken from this network for links within specific functional and locational classifications and applied to the equivalent TIGER classes and link lengths. These classifications were based on grouping roadways according to function (freeway, primary street, secondary street, or other street) and by location inside or outside the downtown area as well as the Portland regional planning boundary (which approximates the urbanized area) (Figure 3).

Figure 3: Average free flow travel times through the Portland street network

    Congested (or peak period) speeds were used for travel between 4:00 PM to 6:00 PM and were calculated in the same manner using the standard Bureau of Public Roads (BPR) speed-flow equation with free flow speeds, link capacity, and peak traffic volumes taken from the local planning network (Dowling Associates 1997).  These congested speeds were transferred to the TIGER network using the same functional and locational classifications as before.  These speeds were used for all travel between 4:00 and 6:00 PM, representing the evening rush hour period in Portland, with free flow speeds used for travel during all other times of the day.  Considerable reductions in driving speeds are apparent due to the application of congested speeds (Figure 4).

Figure 4: Average reduction in travel times due to congestion

    Several space-time accessibility measures, which are based on the concept of the potential path area (PPA), were implemented in this study (Kwan 1998, 1999a; Kwan and Hong 1998; Miller 1991, 1999). This concept can be explained by considering the case of an individual with a daily activity schedule for a number of in-home or out-of-home activities. Some of these activities are considered as fixed in that he or she has little or no control over when and where the activity must take place (such as the workplace). The individual’s mobility is therefore limited by the need to move from the location of the previous fixed activity to the location of the next fixed activity within the time available between these activities. Only the time between successive fixed activities is available for other activities. These include activities such as grocery shopping or filling a car with gas that can be carried out at several possible locations and when convenient, and so can be considered flexible activities.

    The ability to engage in flexible activities therefore depends on the amount of time and mobility available between fixed activities. The area an individual can reach between any two successive fixed activities is the potential path area (PPA). A PPA contains all possible routes an individual could traverse and urban opportunities the individual could potentially reach, given the space-time constraint of the two particular fixed activities in question (Figure 5). When the effect of all successive pairs of fixed activities are considered and their respective PPAs aggregated, these PPAs create a daily potential path area (DPPA) that can be used to assess accessibility for the individual.

Figure 5: Example of Potential Path Area (PPA) (20 minutes duration)

    Because the DPPA is dependent on each individual’s activity schedule, travel through the street network, and the spatial pattern of potential opportunities in the urban area, it can only be found using a dedicated algorithm implemented by GIS-based geocomputational procedures. The spatial and network analytic capabilities of GIS allow not only the measurement of network-based travel times within the context of activity schedules, but also the incorporation of the number and size of potential activity opportunities into the computation. Implementing these procedures using Avenue in the ArcView GIS environment, five space-time accessibility measures were computed. The first is the length of the road segments contained within the DPPA (MILES). The second is the number of opportunities within the DPPA (OPPORTUNITIES). The total area (AREA) and total weighted area (WEIGHTED AREA) of the land parcels within the DPPA is the third and fourth space-time accessibility measures. Finally, to incorporate the effect of business hours on accessibility measures, opportunity parcels were assumed to be available (and could therefore be accessible to an individual) only from 9:00 AM to 6:00 PM.  This creates a fifth accessibility measure, called TIMED AREA. This inclusion of temporal availability adds additional detail to the analysis by emphasizing that while individuals may have considerable constraints on their mobility during the daytime, limited business hours at night will further reduce their accessibility. Both physical mobility and temporal flexibility are therefore necessary to attain high accessibility.

Results and Discussion

    In order to help visualize resulting accessibility patterns, a surface was interpolated for the WEIGHTED AREA measure (Figure 6).  However, it must be remembered that the values of the surface are not dependent on location but rather on individuals who may travel widely throughout the urban area.  This surface shows that individual accessibility is highly variable, with no clear geographic pattern evident within the Portland urban area.  A number of sharp peaks, representing individuals with above average accessibility, are evident in several suburban locations, but there is no peak at or near the CBD.  This is in contrast to the opportunity density surface, which shows that the Portland CBD contains by far the greatest area of potential activities.  Living adjacent to the CBD does not therefore guarantee high accessibility, which is consistent with the findings of space-time accessibility measures that it is individual’s behavior, rather than simply their location, that most strongly influences their accessibility (Kwan 1998).

Figure 6: Weighted opportunity individual accessibility surface for Portland study area

    Individual accessibility in Portland can also be visualized by plotting it as a function of distance from the Portland CBD.  This approximates the standard monocentric urban model and assumes that accessibility should decline with distance from the CBD.  The average accessibility of individuals living within five minute driving time intervals from the CBD is shown in Figure 7.  With the exception of TIMED AREA, the accessibility measures all show a strikingly similar pattern, with access remaining relatively constant until a peak of higher than average accessibility can be seen at a distance of about 20 to 25 minutes driving time.  These are actually the highest accessibility values observed, with individuals possessing values up to 136% of the average.  Beyond this distance accessibility declines, and at the periphery of the city (beyond 35 minutes driving time) the values are far below average.  People living in suburban locations therefore appear to have the highest accessibility, while those on the edge of the city have the least.

Figure 7: Average individual accessibility by distance from the Portland Central Business District (CBD)

    The pattern observed for the TIMED AREA measure is quite different, with the highest values are around ten to 15 minutes distance, and lower than average values found both beyond that distance as well as adjacent to the CBD.  The difference between this measure and WEIGHTED AREA is due solely to behavior, as people living at different locations are engaging in varying amounts of travel during the daytime and so possess greater or lesser access to businesses while they are open.  Because individuals living ten to 15 minutes from the CBD engage in the least amount of nighttime activities they have the least reduction in accessibility when business hours are incorporated, and so now possess higher than average accessibility using this measure (although the absolute values of TIMED AREA is everywhere less than that of WEIGHTED AREA).  Conversely, people living 20 to 35 minutes from the CBD engage in a higher proportion of nighttime activities and suffer a considerable decline in access to opportunities.  Their accessibility as evaluated by the TIMED AREA measure is therefore well below average.  Incorporating time directly into the measure clearly produces a considerably different geography of accessibility in Portland than observed with the other measures.

    The effects of congestion on accessibility can also be shown by distance from the Portland CBD (this time with percent change standardized to a mean of 100, so that values above 100 indicate greater than average reduction, and vice versa).  With the exception of TIMED AREA, the pattern is similar to that for accessibility under free flow conditions (Figure 8).  Reductions are relatively consistent until a suburban peak is reached, followed by below average values on the periphery of the city.  This shows that the effects of congestion are greatest at suburban locations, and are least severe on the edge of the city.  However, the peak reductions are actually at 30 to 35 minutes driving time from the CBD, which is an area of below average accessibility under free flow conditions.  Individuals living at this location therefore tend to have low access to employment and services as well as suffering more from congestion than individuals in other areas, and so are doubly disadvantaged.

Figure 8: Average standardized reduction in individual accessibility by distance from the Portland Central Business District (CBD)

    The TIMED AREA measure tends to exaggerate this pattern, with values for this measure tending to be well above average where WEIGHTED AREA is above average, and vice versa.  Because congestion is applied to travel during the time period that businesses were available, it could be expected that where people tend to do considerable travel during the daytime they would therefore be both more likely to be subject to congestion as well as having greater access to opportunities.  However, this is not the case, as people in these areas (such as ten to 15 minutes from the CBD) actually tend to possess both above average accessibility with the TIMED AREA measure as well as a below average reduction in accessibility due to congestion.  These people are therefore engaging in many activities during the daytime hours but are not suffering much from congestion.  The opposite is also true, so that people living at 30 to 35 minutes distance with relatively few daytime activities may have below average accessibility using the TIMED AREA measure but also suffer above average accessibility reductions during the evening peak traffic period.  Once again, low accessibility is reinforced by severe reductions as a result of congestion.  This is likely occurring because individuals in these areas are driving on very congested roadways and so are experiencing considerable congestion during the relatively few activities that they take part in during the evening rush hour.  It is not just when they carry out the activities but where they travel that influences the severity of the congestion they face.

    The importance of distance to individual accessibility can also be assessed within a polycentric framework, using 12 regional centers defined by the local planning agency in their future growth plan for Portland (Metro 1997).  These centers include the Portland CBD, the downtowns of several suburbs, major shopping centers, and suburban employment concentrations.  As with the monocentric model, distance again should determine accessibility, but this time from multiple points. Because these centers (which include the Portland CBD) are so widely distributed there are fewer distance intervals, but strong patterns can nonetheless be observed (Figure 9).  For all but the TIMED AREA measure accessibility is highest close to centers and lowest at farthest distances, though there is not a direct relationship between distance and access.  This pattern is seen more strongly with TIMED AREA, especially the very high accessibility values adjacent to the centers (though the absolute values of this measure were on average 65.7% lower than WEIGHTED AREA).  These high values are again due to behavior, with individuals adjacent to the regional centers engaging in a higher proportion of their activities during the daytime than people living farther away.

Figure 9: Average individual accessibility by distance from twelve regional centers in the Portland metro area

    When congestion is applied reductions in accessibility are very even except at the farthest distances, where very high reductions are observed (Figure 10).  This pattern is true of all measures, though the TIMED AREA measure shows the greatest reductions at the farthest distances. This is interesting as it is at these locations that individuals engage in a lower proportion of daytime activities, and so would not be expected to be subject to reductions as a result of daytime congestion.  However, as with the monocentric model, it may also be that these individuals live or move about in areas of the city with highly congested streets, and so suffer considerably from congestion during the daytime activities they do engage in.  And again, those individuals possessing the least accessibility also suffer the most from congestion.  Taking part in few daytime activities may indeed be a response to the congestion they face during the daytime.

Figure 10: Average standardized reduction in individual accessibility by distance from twelve regional centers in the Portland metro area

Conclusions

    The incorporation of time into the evaluation of individual accessibility within cities has produced interesting and sometimes unexpected results.  Link-specific travel times produce very uneven accessibility patterns, with access to services and employment varying considerably within Portland.  The time of day activities were carried out has also been shown to have an effect on accessibility, as evening congestion sharply reduced individual’s access throughout the city.  The effects of this congestion on mobility is highly spatially uneven, even though in this study congested traffic speeds were only applied to a two hour time period in the evening.  More temporally precise applications of congestion (perhaps capturing hourly variations) would therefore likely reduce accessibility for all individuals to a much greater extent than has been observed in this study, and for some people and areas more than others.  The use of business hours to limit access to opportunities at certain times of the day with the TIMED AREA variable has shown that time of day can be incorporated into accessibility measures, and shows that non-temporally restricted accessibility measures are producing inflated values by treating these opportunities as being available at all times of the day.  It is not just that incorporating time reduces accessibility, but that it also produces a very different, and perhaps unexpected, geography of accessibility.  This geography depends much on individual behavior and so cannot be discerned from the location of opportunities or congestion alone.

    Another important finding of this research is the fact that the use of greater spatial and temporal detail in accessibility measures produces patterns that do not fully support monocentric and polycentric notions of urban form and human behavior.  With the monocentric model people living in suburban locations tend to have the highest accessibility (despite the observed CBD peak of opportunities), while those living at the periphery of the city have the lowest.  Areas closer to the CBD, which Abbott (1983) has characterized in a monocentric discussion of Portland as including several distinct types of pre-automobile neighborhoods, show more even accessibility patterns (except for TIMED AREA, which incorporates the importance of the time of day that activities are carried out).  Reductions due to congestion are biased against more recently developed suburban areas, perhaps reflecting the inability of street improvements to keep up with traffic and the presence of heavy cross-commuting patterns within suburban areas.  Stronger evidence perhaps exists in favor of the polycentric model, which is rather surprising given the limitations of this model in capturing human behavior in other urban contexts (Pickus and Gober 1988; Hoch and Waddell 1993; Waddell, Berry, and Hoch 1993; Fujii and Hartshorn 1995).  However, the TIMED AREA measure again varies from this pattern by greatly increasing the accessibility of those adjacent to centers, while the effects of congestion also fail to support this model as they are very uneven with distance.  So even while it does appear to provide some explanation for accessibility patterns in Portland, the fact remains that there are significant differences from the behavior expected of individuals by these models, due in large part to the role of time within the accessibility measures.  Instead, the results here confirm findings by other space-time accessibility research that these measures do not reflect proximity to features within the urban environment but instead are based on individual’s own experiences of the city (Kwan 1998).  This is significant, because while the importance and role of distance in influencing human behavior and land use in these models has been strongly questioned by recent evidence (Giuliano 1989, 1995), their influence is nonetheless still common in a variety of urban applications.

    These issues are also important because work on gender, commuting and accessibility, as well as the very existence of congestion, has shown that much of daily travel is carried out at particular times of the day, especially in the evening (Kwan 1999b).  The accurate modeling of congestion resulting from this travel activity scheduling, can be expected to be crucial to realistically evaluating accessibility.  This is especially the case because the amount of travel and its temporal scheduling has been shown to vary by gender and employment status (Blumen 1994; Kwan 1999b).  This means that congestion will almost certainly affect men and women, as well as part or full time workers, differently.  And while this study has focused on those traveling exclusively by personal auto, the importance of time to individual’s access to employment or services will likely be even greater for those dependent on public transit, for whom mobility is dependent not only on their own constraints but the scheduling of transit systems (which will vary by time of day and day of week).

Acknowledgment

This paper is an abbreviated version of a paper currently under review for publication

References

Abbott, Carl. 1983.  Portland: Planning, Politics, and Growth in a Twentieth-Century City.  Lincoln: University of Nebraska Press.

Black, J., and M. Conroy. 1977. Accessibility measures and the social evaluation of urban structure.  Environment and Planning A 9:1013-1031.

Blumen, Orna. 1994. Gender differences in the journey to work. Urban Geography 15: 223-245.

Brainard, Juli, Andrew Lovett, and Ian Bateman. 1997. Using isochrone surfaces in travel cost models. Journal of Transport Geography 5: 117-126.

----. 1999. Integrating geographical information systems into travel cost analysis and benefit transfer. International Journal of Geographical Information Science 13: 227-246.

Cambridge Systematics, Inc. 1996. Data Collection in the Portland, Oregon Metropolitan Area Case Study. Washington, DC: US Department of Transportation.

Dowling Associates. 1997. Travel Model Speed Estimation and Post Processing Methods for Air Quality Analysis. Washington, DC: United States Department of Transportation.

Fujii, T., and Truman A. Hartshorn. 1995. The changing metropolitan structure of Atlanta, Georgia: locations of functions and regional structure in a multinucleated urban area. Urban Geography 16: 680-707.

Geertman, Stan C.M., and Jan R. Ritsema Van Eck. 1995. GIS and models of accessibility potential: an application in planning. International Journal of Geographical Information Systems 9: 67?80.

Giuliano, Genevieve. 1989. New directions for understanding transportation and land use. Environment and Planning A 21: 145-159.

----. 1995. The weakening transportation-land use connection. Access 6: 3-11.

----. 1993. Regional versus local accessibility: implications for nonwork travel. Transportation Research Record 1400: 58-66.

Helling, Amy. 1998. Changing intra-metropolitan accessibility in the U.S.: evidence from Atlanta. Progress in Planning 49: 55-108.

Hoch, Irving, and Paul Waddell. 1993. Apartment rents: another challenge to the monocentric model. Geographical Analysis 25: 20-31.

Knox, Paul L. 1978. The intraurban ecology of primary medical care: patterns of accessibility and their policy implications. Environment and Planning A 10: 415-435.

Kwan, Mei-Po. 1998. Space-time and integral measures of individual accessibility: a comparative analysis using a point-based framework. Geographical Analysis 30:191-217.

----. 1999a. Gender and individual access to urban opportunities: a study using space-time measures. Professional Geographer 51: 210-227.

----.1999b. Gender, the home-work link, and space-time patterns of nonemployment activities. Economic Geography 75: 370-394.

Kwan, Mei-Po, and Xiao-Dong Hong. 1998.Network-based constraints-oriented choice set formation using GIS. Geographical Systems 5:139-162.

Lenntorp, Bo. 1976. Paths in Space-Time Environments: a Time-Geographic Study of the Movement Possibilities of Individuals. Lund Studies in Geography B: Human Geography. Lund: Gleerup.

----.1978. A time-geographic simulation model of individual activity programmes. In Human Activity and Time Geography, ed. Tommy Carlstein, Don Parkes, and Nigel Thrift. London: Edward Arnold.

Metro. 1997. Regional Framework Plan. Portland: Metropolitan Service District.

Miller, Harvey J. 1991. Modelling accessibility using space-time prism concepts within geographic information systems. International Journal of Geographic Information Systems 5: 287-301.

----. 1999. Measuring space-time accessibility benefits within transportation networks: basic theory and computational procedures. Geographical Analysis 31: 1-26.

Miller, Roger. 1982. Household activity patterns in nineteenth-century suburbs: a time-geographic exploration. Annals of the Association of American Geographers 72: 355-371.

Muraco, William A. 1972. Intraurban accessibility. Economic Geography 48: 388-405.

Pickus, J., and P. Gober. 1988. Urban villages and activity patterns in Phoenix. Urban Geography 9: 85-97.

Scott, Lauren M. 1998. Evaluating intra-metropolitan accessibility in the information age: operational issues, objectives, and implementation. Paper presented at Varenius conference, November 20-22, 1998, Pacific Grove, CA.

Wachs, Martin, and Gordon T. Kumagai. 1973. Physical accessibility as a social indicator. Socio-Economic Planning Science 7: 437-456.

Waddell, Paul, Brian J.L. Berry, and Irving Hoch. 1993. Housing price gradients: the intersection of space and built form. Geographical Analysis 25: 5-19.

Wang, Fahui. 2000. Modeling commuting patterns in Chicago in a GIS environment: a job-accessibility perspective. Professional Geographer 52: 120-133.