Michael Gould
This paper discusses ideas on past, present and possible future ways to model accessibility using geographic information technologies. Although accessibility modeling is traditionally associated with shortest path (transit) calculations, we have used it for urban and regional planning purposes to address not so much the routes as the population or area served (or marginalized) by possible routes. Third Wave phenomena such as telecommuting and teleshopping are changing the way we view and model urban accessibility. Primary interest rests on developing and mapping new indices to represent the concept of virtual accessibility to economic activity centers.
Accessibility
The study of accessibility began in the 1850s (Carey 1858-9) and was popularized during the quantitative revolution a century later, beginning with economists like Stewart, Zipf and Huff. Most early attempts to represent spatial opportunity (of cities, individual shoppers, etc.) were based on aggregate accessibility and were purely topological: nodal connectivity (after Shimbel 1953), alpha and beta indices (Kansky 1963) and general use of connectivity (O-D) matrices without any sort of spatial technologies including cartography. More complex, geographic accessibility models then went beyond topology to consider along-network distances, weighted links and nodes to simulate attraction, intervening opportunity, etc., resulting in a variety of useful disaggregate accessibility indicators or measures of how easily people get the services they need.
One example of such a geographic accessibility indicator (GAI) was developed for a study of transport accessibility in Spain, sponsored by the Ministry of Public Works and Transport (Gutiérrez, García, Gould and Monzón 1993). The GAI captures the notion of infrastructure quality but moreso the geographic centrality of each origin node (i), so that large cities located near the centre of a region receive large accessibility values while peripheral ones (of the same size) receive lower values. Given a point pattern (450 populated places in our national study; at an urban scale they might be shoppers, school children, medical patients, etc.), we calculate the overall opportunity of each point, as follows:
where
RI is the real impedance via the network, between nodes i and j, and ICEA is the Income (or attraction) of the destination Center of Economic Activity (j).
The GAI is scale independent, and was also applied to the Metropolitan Madrid street network (see map 4, by Gómez) to visualize accessibility taking into consideration each of the M-30, M-40 and future M-50 orbitals. Another Indicator of Accessibility by Infrastructure (IAI), focuses primarily on the quality of the transport network connecting i and j:
where
RI is the real, calculated impedance via the network, between nodes i and j, ICEA is the Income (or attraction) of the destination Center of Economic Activity (j), and II is the Ideal Impedance between i and j (assumes Euclidean distance and optimal infrastructure).
The IAI helps us to understand the situation whereby people may live geographically near to an airport or the CBD, for instance, yet have poor accessibility due to road conditions, traffic, and other attributes of the real-world, physical infrastructure.
In the case of both indicators the Real Impedance, for the case of highways, is simply the summation of arc (link) and node impedances, as follows.
The most important impedance, or travel cost, in our study was the sum of travel frictions along arcs, most importantly network distance but also capacity (highway class), average speeds, percentage heavy vehicles, etc. of each arc, as follows:
where
T is the expected travel time, based on published distances between i and j and the speed limits of the link(s),
Ci is a coefficient of infrastructure, where highways are given a small negative friction, average provincial roads zero friction and lesser quality roads a small positive friction (penalty),
Ct is a coefficient of traffic, considering factors such as average traffic intensity and percent heavy vehicles, and
Cn is a coefficient of nationality, whereby links crossing between Spain and neighboring regions in Portugal and France receive a slight friction or penalty.
Node impedances were not considered in detail in the national study, but are routinely used in urban network modeling because they may represent factors involving traffic intersections, turn restrictions or penalties, as well as simple crossing penalties (important to the discussion below).
The values of both GAI and IAI are calculated using the network modeling capabilities, such as determination of shortest paths from i to j, of a feature-based GIS. These values become ordinary point attributes in the GIS database. Once each point of interest (city, household, etc.) receives its accessibility value we then interpolate isoaccessibility lines, determine area or population falling within each accessibility region, and model temporal sequences of accessibility growth or decline. An example for Spain is found in maps 1-3, which show differences between IAI values in 1992 and those expected in 2007 if the Ministry's Infrastructure Plan is fully implemented. Variations on GIS-based accessibility modeling in an urban context are found in Lakerveld (1992), Arentze, Borgers and Timmermans (1992), Geertman and Reitsema van Eck (1995) and Gutiérrez, Gómez and Gould (1996). The use of geographic information technologies has allowed accessibility modeling to move from the sterile, isotropic world of O-D matrices to more realistic geographic networks and to provide more descriptive output using automated cartography. Furthermore, using geographic information technologies accessibility can more easily be linked to, or compared with, other geographic themes such as land use and property value (cadastre) for predictive modeling. Thus, automated accessibility analyses can provide the urban or regional planner with more scenarios more quickly, each the synthesis of relevant spatial information (normally layers).
Third Wave Accessibility
The Toffler's (1980; 1995) First-Wave civilization saw accessibility in
terms of food gathering and production: people generally settled near
fertile areas. The Second Wave treats accessibility in terms of
transportation of raw materials to processing facilities, and then to the
customer. Recently, the sweeping diffusion of telecommunication
infrastructure and services has spurred a new view of accessibility, so
that it can no longer be described strictly in terms of spatial
opportunity. Third Wave accessibility goes beyond physical access to
tertiary economic centers such as shopping malls or medical centers, to
describe virtual access to quaternary/quinary centers: information
services. This phenomenon was distinguishable some 20 years ago in the USA
when the telephone company convinced millions of Americans to "let their
fingers do the walking".
Access to information on regional and personal preferences and lifestyles
led to market segmentation, which fueled the micromarketing revolution,
including shopping from home via printed catalogs and television offers,
using the ubiquitous credit card. Southern Europeans have not succumb
(yet) to this information immersion and subsequent marketing blitz,
preferring the intense personal contact of daily street life; dealing at
times with large amounts of cash and waiting in lines is a price many are
still willing to pay. This is a clear trait of Second Wave society, on the
verge of entering the Third. The concept of accessibility in the American
city began evolving more than a decade ago when commerce moved from the
street to telephone and express courier networks. The latter peaked during
the late 1980s, but is already admitting heavy casualties at the hands of
fax and E-mail (and the US Postal Service expects to lose 50% of its
traditional business by 2000): "next day" is too long to wait.
Under the Third Wave model of accessibility traditional geoeconomic models
break down, because the spatial dimension (distance, centrality) is
essentially lost. The toll-free phone number now permits that someone in
Maine wishing to purchase a computer can just as easily buy it from San
Diego than from Bangor. The only impedance in the virtual shopping network
is temporal: a courier may require two days instead of a half day to make
delivery. Of course this assumes we are buying atoms not bits. Pushing
the virtual shopping network a bit further, we find financial institutions
transferring bits representing billions of dollars from New York to Hong
Kong with the click of a mouse. Transmission along high-speed networks is
theoretically the speed of light, but is practically constrained by network
traffic at the many intersections: essentially the node-crossing penalty
mentioned above. Thus, given equal network infrastructure and protocols,
the difference between an interurban and a transcontinental electronic
transfer has little to do with distance per se, but rather with the number
of nodes (switches) crossed along the way. So we are back to Shimbel's
index of nodal connectivity! Curiously, like many spatial relations in our
everyday (naive) geography including those used for interurban navigation,
topology without geometry proves sufficient.
Where, then, to locate a Third Wave business? Again, Christaller and Lösch
are brushed aside as space gives way to cyberspace. Traditional models of
urban accessibility are simply not relevant today. At urban scales, a
financial institution need not be located on Wall Street ( a prestige P.O.
box perhaps), because location on a high-bandwidth network-anywhere in the
city or the world, really- is all that is necessary. Today's
information-based companies are taking advantage of Third Wave
accessibility by relocating quality control facilities to locations around
the globe which are convenient with respect to information production.
American software companies are opening QC outposts in small, tranquil
villages in Hawaii and Scotland, offering employees a stress-free lifestyle
and the company a 6 hour differential. Information (databases, source
code) is uploaded at quitting time on the east coast, when in Hawaii it is
time to start work. The next morning EST, the corrected information has
been downloaded from Hawaii and is waiting to go. In the case of
Internet-based services, some of the more interesting and heavily utilized
WWW services are located in remote areas having no substantial geographic
market. Mirroring of important databases across spatially-distributed
servers assures users in whichever region minimum response time.
The Digital City
New infrastructure, whether technologically new (high bandwidth telecom
networks) or just additional (more highways), is changing the composition
of the Third Wave city. At the onset of the Third Wave in the USA (1960s),
edge cities were the result of the desire to escape the congested,
expensive central city. Rent was cheaper on the fringe and there was no
need for information and high-tech service industries to be located near to
raw materials or cheap labor in the CBD: many high-tech employees lived in
the suburbs already. Landuse and land values changed, as the suburban
region or belt became home to enormous, tax-paying R&D and commercial
parks. Now that the Third Wave is well established (some would say already
burning out) companies now realize that physical location adjacent to the
large city is less important than ever. Thus, the edge city is losing
traditional tenants and property values are declining in an effort to
attract newcomers. This inevitably affects land use in major cities, a
trend less noticeable in places where cities are smaller such as in
southern Europe, though it is soon expected to reach such areas according
to predictions of 200 experts of the Europe 2000 project (Hall, 1976).
Specifically, this study predicted that in the year 2000 many workers in
the Madrid area will live in tranquil farmhouses between 70 and 150 km from
the capital, in the Sierra de Guadarrama for example. It is interesting to
note that this Sierra is now growing very quickly in popularity, not only
for second residences but also for primary ones. Unfortunately, most are
long-range (traditional) commuters who drive more than an hour to work each
day: the mountain roads which empty onto the Madrid plain are the cause of
some of the worst traffic jams in the city. This may change soon, however,
thanks to fierce competition among Internet Service Providers (ISP) and a
special Internet rate offered by the phone monopoly, making it possible to
be on-line all day (8 hours) for as little as 1200 pesetas ($9.00)...about
the cost of gasoline necessary to commute 100 km round trip.
Implementation Possibilities
How can geographic information technologies help urban planners and
analysts? In the area of accessibility, GIS can be used to implement and
visualize the results of new models (indices) which synthesize physical and
virtual attributes of accessibility. It allows us to ask what are the
geographical effects of activities essentially devoid of geographical
dimension, as are telecommuting and teleshopping. In the same way we
presently calculate accessibility via road or rail, we hope to add virtual
accessibility to the formula, and then search for inequalities: determine
which connected populations become better connected and which marginalized
populations become further removed from society's opportunities.
The product of the accessibility indices mentioned above is a single value
assigned to each node of interest (populated places, customers, etc.).
Therefore, just as several economic and network-based attributes figure in
the final accessibility value, we may add virtual accessibility (VA)
attributes as well. Candidates for VA attributes are the following:
Scenario 1. Person A lives in the center of a primate city; in an
intelligent building wired with a T-1 connection; has direct access to
Internet; is fluent in English; runs a home business (stock trading) via
the Net; walking distance to major economic activity.
Scenario 2. Person B lives 80 km from a secondary city; in a farmhouse in
the hills; not near to any major highways; connects to Internet via POTS
(14.4 Kbps); understands some English but relies mostly on spanish servers.
Scenario 3. Person C lives in a isolated rural village of 200 inhabitants;
one public phone in town; 120 km from major city; travels to city once or
twice annually.
The key to determining the GAI and IAI described earlier, is the Real
Impedance (RI), whose increases causes accessibility to decrease. Just as
our coefficient of infrastructure, Ci, uses negative impedances for
high-speed motorways (or TGV-class trains), so too may negative attributes
be associated with virtual connectedness. The summation of arc and node
impedances, therefore, is the logical place to add VA attributes. An
optimal scheme has yet to be worked out (suggestions solicited at the
Conference), but it seems evident that basic telephone connection is a
binary attribute which should enter into the equation. Actual
telecommunication services available and exploited, however, is a difficult
issue to handle. In the case of Spain, anyone with a telephone connection
can, theorectically, connect to Internet via the telephone company's
recently unveiled nationwide InfoVía service. Perhaps a better indicator
of potential connectedness, then, is overall purchasing power, necessary to
assume the cost of a PC and modem, along with the minimal (but real)
connection (telephone) charges. This cost, on the order of $2000 fixed and
$30/month variable, cannot be assumed by everyone, no matter how appealing
the Internet offering may be. (It would be fascinating to be able to look
20 years into the future, to see what might happen if Newt Gingrich's idea
of donating laptop computers and modems to America's poor were put into
practice!)
Considering these possible attributes in terms of reducing the impedance to
accessibility, therefore, would change the accessibility maps shown
earlier. In the case of cities (neighborhoods, blocks, buildings) having a
high number of people as in scenario 1, the overall accessibility level
would increase, effectively "nearing" these people to the economic services
(information) they need. In the case of scenario 3, the already low
accessibility level-based on lack of road connections-would be reduced
still.
We are hoping to receive funding from the telephone monopoly (Telefónica)
and the newly created Ministry of Development, to update our urban,
regional, and national accessibility maps, to include notions of virtual
accessibility.
Acknowledgements
The studies deriving and utilizing GAI and IAI were directed by Javier
Gutiérrez Puebla of the Univ. Complutense de Madrid. The research
currently falls under general support by the Extremadura regional
government (Junta de Extremadura, Consejería de Educación y Juventud).
Bibliography
Arentze, T., Borgers, A., and Timmermans, H., 1992, Geographical
information systems, accessibility and multipurpose travel: A new
measurement approach. Proceedings of the Third European Conference on
Geographical Information Systems (EGIS 92) held on March 23-26 in Munich
(Utrecht: EGIS Bureau), 438-450.
Carey, H.C., 1858-9, Principles of Social Science, Philadelphia: Lippincott.
Eskelinen, H., 1993, Rural areas in the high mobility communications
society. In Transport and Communications Innovation in Europe, edited by G.
Giannopoulos and A. Gillespie (London: Belhaven).
Geertman, S.C.M., and Reitsema van Eck, J.R., 1995, GIS and models of
accessibility potential: An application in planning. International Journal
of Geographical Information Systems 9, 67.
Gould, M., 1996, Nuestro futuro decentralizado, Mapping (Madrid), July
1996, in press.
Gould, M.D. and Gutiérrez, J., Regional Infrastructure-based Accessibility
Calculation using GIS, manuscript under consideration.
Gutiérrez, J., Gómez, G. and Gould, M., 1996, Where do we go from here?:
Planning Transport Infrastructure at European Regional Level, GIS Europe,
5(4), xv-xvii.
Gutiérrez, J., García, S., Gould, M., and Monzón, A., 1993, Implementation
of a national motorway accessibility model. Proceedings of the Fourth
European Conference on Geographical Information Systems (EGIS 93) held in
Genoa, Italy on 29 March-1 April 1993 (Utrecht: EGIS Foundation),
1373-1380.
Kansky, K. J., 1963, Structure of Transportation, Research Paper 84,
Department of Geography, University of Chicago, Chicago, Illinois.
(photocopy provided by D.F. Marble)
Hall, P. (ed.), 1976. Europe 2000, London: Duckworth.
Lakerveld, M., 1992, Allocation modeling at municipal level: Putting GIS
into practice. Proceedings of the Third European Conference on
Geographical Information Systems (EGIS 92) held on March 23-26 in Munich
(Utrecht: EGIS Bureau), 92-102.
Negroponte, N., 1995. Being Digital, New York: Alfred Knoft.
Shimbel, A., 1953, Structural parameters of communication networks.
Bulletin of Mathematical Biophysics 15, 501.
Toffler, A., 1980. The Third Wave, New York: Bantam.
Toffler, A. and H. Toffler, 1995. Creating a new civilization: The politics
of the Third Wave, Atlanta: Turner Publishing.
Map 1
Map 2
Map 3
Map 4
One can easily imagine the approximate virtual accessibility levels of the
people in the following scenarios.
Michael Gould
Departamento de Geografía y Ordenación del Territorio
Universidad de Extremadura
E-10004 Cáceres, Spain
Fax: (+34) 27 24 88 58
Email: mgould@unex.es
http://geot.unex.es/english/fichas/~mgould.html