Accepted Manuscript Title: Mapping accessibility to generic services in Europe: a market-potential based approach Authors: Mert Kompil, Chris Jacobs-Crisioni, Lewis Dijkstra, Carlo Lavalle PII: DOI: Reference:
S2210-6707(18)31263-0 https://doi.org/10.1016/j.scs.2018.11.047 SCS 1372
To appear in: Received date: Revised date: Accepted date:
29 June 2018 30 November 2018 30 November 2018
Please cite this article as: Kompil M, Jacobs-Crisioni C, Dijkstra L, Lavalle C, Mapping accessibility to generic services in Europe: a market-potential based approach, Sustainable Cities and Society (2018), https://doi.org/10.1016/j.scs.2018.11.047 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Mapping accessibility to generic services in Europe: a market-potential based approach Mert Kompila,b, Chris Jacobs-Crisionib, Lewis Dijkstrac and Carlo Lavalleb a
SC RI PT
Corresponding author – e-mail:
[email protected] European Commission, Joint Research Centre (JRC), Ispra, Italy c European Commission, DG for Regional and Urban Policy, Brussels, Belgium b
Highlights
This study introduces a universal method to measure access to ‘generic’ services using a detailed population distribution and transport network.
It manages to reproduce observed service accessibility patterns with reasonable accuracy.
Model results show that cities provide better opportunities in accessing services and there is a big gap
N
U
A
between urban and rural areas, especially in access to regional services. It has been estimated that four out of five people in Europe have at least one local or daily service
M
within 5 kilometres. On the other hand, based on the model results, one out of four people in Europe lack a regional or high-order service within 40 kilometres.
D
Cities offer far more potential to access services by walking and cycling.
TE
EP
Abstract
An important goal of land use and transport policies is improving accessibility to services. Access to services
CC
differs significantly between different territories, and regional development policies, including Cohesion Policy, are often used to improve accessibility where it is too low. Unfortunately, comprehensive data on facility locations is not available in Europe, making it difficult to evaluate, ex-ante, the direct or indirect effects of
A
policies on service distribution and accessibility to services. This study proposes a novel approach to cope with this problem and maps generic service accessibility in Europe. First, it defines three types of generic services with different minimum number of users and characteristics: local, subregional and regional services. Next, it uses this to simulate the spatial distribution of services, and to assess probable accessibility to services across Europe. Based on the simulation results, the paper explores how per country accessibility to services differs between metro regions, and between urban and rural areas in terms of I) average distance to services and II) 1
the share of population within a short enough distance that could be walked or cycled. It tests the proposed method by comparing actual service points with the modelled points in a selected number of countries.
Keywords: accessibility, accessibility to services and facilities, proximity, population grid, degree of urbanisation,
SC RI PT
cities, metropolitan regions, rural areas.
1. Introduction
Today, cities are associated with major environmental, social and economic challenges. Air and water pollution, traffic congestion, social segregation, poverty, inadequacy of social services and physical
infrastructure, and insufficiency in mobility and accessibility are among the most important ones. To cope with
U
those challenges, cities and regions implement sustainable policies, in particular, to maintain economic
productivity, improve the quality of life and address other environmental and social problems. The spatial
N
distribution of activities, efficient use of resources and accessibility to different services and facilities are crucial
A
to promote more sustainable cities and regions in terms of efficient land use, urban form, other operations and
M
functions (Bibri and Krogstie, 2017; Bourdic et al., 2012). In the transport sector, policies promoting sustainability usually focus on reducing negative effects, such as high energy use, pollution and emissions, and
D
providing a high level of accessibility to individuals and organizations. Several policies are common including: 1) enhancing accessibility rather than increasing mobility (i.e. traffic); 2) integrating transport and land use
TE
planning; 3) creating mixed-use development in close proximity to public transport; and 4) encouraging walking and cycling (Curtis,2008; Bertolini, 2005; Banister, 2008; Ferreira et al., 2012).
EP
Transport and accessibility in cities are emphasized by the United Nations’ (UN) Sustainable Development Goals (SDGs). The UN (2018) specific targets for sustainable cities and communities include providing access to
CC
safe, affordable, accessible and sustainable transport systems for all, expanding public transport in cities, enhancing inclusive and sustainable urbanization and providing universal access to safe, inclusive and accessible, green and public spaces. In parallel to the UN’s SDGs, the European Union under the framework of
A
Urban Agenda for the EU (EC, 2018a; 2018b) works with cities and regions to develop a sustainable urban mobility policy including efficient public transport systems and by promoting active mobility solutions, such as walking and cycling, and by ensuring good accessibility for residents and commuters. For the SDGs and the Urban Agenda for the EU, accessibility to services, together with the land use and mobility interconnections feature prominently among the policies for more sustainable cities.
2
Accessibility can be broadly defined as the degree in which relevant destinations can be reached given available transport means. It is increasingly considered a key policy goal in land-use and transport planning and has been promoted as the most relevant criterion in policy evaluations. Improving accessibility to services is an important part of spatial plans and policies across Europe. An improved accessibility to services can reduce social and territorial disparities and decrease transport costs for the users. Unfortunately, comprehensive data on facility locations is not available in Europe, making it difficult to evaluate, ex-ante, the direct or indirect
SC RI PT
effects of policies on service distribution and accessibility to services. As it would be necessary for policy
evaluations, harmonized and comprehensive data on service locations is necessary. Unfortunately, such data is unavailable or incomplete, and hence, knowledge about accessibility to services in the EU territory is limited. This study proposes a novel approach to cope with this problem and maps generic (‘generated’ or ‘general purpose’) service accessibility in Europe. At first, it defines three types of generic services based on their
minimum number of users: local, subregional and regional services. Then it uses this definition to estimate the
U
spatial distribution of services and, finally, to assess probable accessibility to services across the EU. It applies a
N
method that is based on the local market potential (i.e., potential users or clients) and aims to find answers to
A
the following questions:
How is accessibility to generic service areas and facilities structured in Europe?
How does accessibility to services differ among countries and regions and to what extent can these
M
differences be explained by different patterns of population distribution? Considering accessibility to generic services, how large is the gap between urban and rural areas.
What percentage of people may be able to walk or cycle to their services? How does that differ among
TE
D
different services, e.g. at local or regional level?
EP
To answer those questions, an accessibility indicator has been developed with the following operational features in principle: 1) easy to understand and implement, 2) universal across various types of facilities and among countries, 3) responsive to changes in population distribution and in transport service quality, and
CC
finally, 4) capable of simulating (partly) supply and demand side changes through parameter changes. The following section covers a literature review on accessibility and access to services. Section 3 elaborates the
A
methodology and the data behind the developed indicator. Section 4 describes the results per country across the EU for different types of areas and regions. Section 5 compares the simulated results with selected observations for validation purposes. Finally, section 6 evaluates the findings of the study and discusses potential policy implications.
3
2. Literature review There are many methods to measure accessibility; see for example Geurs and van Wee (2004), Gutiérrez (2009), Curtis and Scheurer (2010), and Páez et al. (2012). Those measures mostly share a dependency on the spatial distribution of the studied destinations, their attractiveness, and the performance of the available transport system. In that sense, accessibility measures differ substantially from typical transport infrastructure
SC RI PT
evaluations and are particularly useful in evaluating the interaction between spatial activities and transport networks.
In line with the diversity of available accessibility measures, accessibility related policy goals differ substantially. For instance, accessibility may play a role in policies that aim to improve regional cohesion (López et al., 2008; Stepniak and Rosik, 2013), incite economic growth (Keeble et al., 1982; Vickerman et al., 1999), or reduce social exclusion (Lucas, 2012; Lucas et al., 2016). Those different aims are often tied to different ethical stances, such
U
as utilitarianism, libertarianism or egalitarianism (van Wee and Geurs, 2011; Pereira et al., 2017).
N
At the European Union (EU) level, a utilitarian stance has long prevailed, as witnessed by the long-lasting efforts to promote aggregate regional growth through cohesion policies and transport infrastructure
A
improvements (European Commission, 2004; Jacobs-Crisioni et al., 2016; Keeble et al., 1982; López et al., 2008;
M
Spiekermann et al., 2015; Vickerman et al., 1999). However, recent years have brought new challenges for Europe’s territory, which give rise to questions regarding the fairness and equity aspects of EU policies: for
D
instance, to what degree can residents access basic services and how could current policies help to improve the
TE
availability and accessibility of those services? This is particularly relevant for some of Europe’s regions, where a highly dispersed population, ageing and population decline contribute to market conditions that are unfavourable for the cost effective provision of services. The provision of fair and balanced accessibility to
EP
services can be seen a key aspect of reducing territorial disparities across Europe. To measure accessibility to local and regional services, many researchers developed measures based on travel
CC
time or distance. Access to healthcare and travel time to hospitals are repeatedly studied, as are those for effective emergency treatment (Nicholl et al., 2007; McGrail, 2012; Pilkington et al., 2017), often leading to
A
travel-related concerns in the context of rural hospital closures or shortages in provision of hospitals. Travel distance to schools is another frequently studied area, which is important due to its links with the choice for active transport modes and, consequently, children’s health (McDonald, 2008). Distance to shops and cultural facilities has equally been found an important factor in studies concerning for example food deserts and obesity (Ghosh-Dastidar et al., 2014), social exclusion of the elderly (Hirai et al., 2015) and city attractiveness (Oner, 2017; Garretsen and Marlet, 2017). 4
Service accessibility requires a very different approach than the utilitarian-oriented methods common in EU evaluation practice, requiring a much more local perspective than is customary. A recent ESPON project at European level explored to what extent the varying levels of Services of General Interest (SeGI) contribute to the competitiveness, economic development and job growth of different territories (Rauhut et al. 2015; Breuer et al., 2013; Fassmann et al., 2015). The project developed various indicators to explore availability, adequacy and provision of services varying from gas and electricity to telecommunication infrastructure or from labour
SC RI PT
market services or hospitals to physicians and hospital beds. Marques da Costa et al. (2015) discussed regional disparities of SeGI provision with the analysis of transport and communication infrastructure, education and health services. Similarly, Rauhut and Komornicki (2015) analysed SeGI provision in rural areas and explored centrality of services in rural and urban contexts. However, those studies unfortunately do not include or
elaborate spatial / physical accessibility to services. Differently, Milbert et al. (2013) presented comparative results on accessibility of SeGI across five different regions in Europe. They explored accessibility to low
U
(primary schools), medium (railway stations) and high centrality (airports) services using high-resolution spatial
N
data. In another study, Papaioannou and Wagner (2017) have developed location based accessibility indicators for measuring accessibility to the closest school and hospital weighted by population. They reported
A
comparable results in the form of travel times and speeds for private cars and public transport in 18 globally
M
selected cities. Finally, another recent project called PROFECY (ESPON, 2017a; 2017b) used accessibility measurements to identify inner peripheries, national territories facing challenges of access to basic services of
D
general interest. In this project, it was identified European regions with poor accessibility to regional (economic and demographic) centres and various services of general interests such as health care and education, banks,
TE
cinemas and train stations. The data was derived mainly from the Open Street Map and accessibility was measured via travel time by car.
EP
Considering the reviewed studies, with the method for measuring accessibility to generic services, this study intends to contribute in (1) extending service accessibility measures to a pan-European level, where
CC
information on service area locations are usually incomplete or inadequate; (2) simulating EU-wide accessibility to generic services in fine spatial resolution, which facilitates international / interregional comparisons for
A
policy analysis (also for urban and rural areas); and (3) introducing a policy tool for monitoring potential impacts of population and transport infrastructure changes on accessibility to services; or for comparing real landscapes with idealized landscapes to highlight overserved and underserved areas by a specific service.
5
3. Methodology and data 3.1. Background information The generic services modelled in this study are essentially proxies for any type of public or private service. They provide goods or services directly to a region’s residents and require physical presence of the resident at the service location, such as education and health facilities, childcare services, cultural and sports facilities,
SC RI PT
commercial areas for food or goods. Models for the location choices of facilities as actors have been proposed repeatedly, emphasizing mainly the roles of centrality and agglomeration. Von Thünen et al. (1826) suggested that, everything else held equal, transport costs to a central market are the main organizing factor in the spatial distribution of farm activities. In his central place theory, Christaller (1933) proposed a system of settlements with a characteristic hierarchy of offered services, based on the catchment area the presumed frequency of use of those services. All settlements would offer services with a small catchment area that are typically needed at
U
a daily basis. Only one most central settlement would offer services that are needed by all residents of the
N
system, but not often enough to enable a finer distribution of that service through the system, so that the catchment area of those services would entail the entire system. Lösch (1938; 1954) expanded Christaller’s
A
work with a bottom-up hierarchical approach where central functions finally form a regularly distributed
M
triangular-hexagonal pattern, in other words, an ideal landscape of central functions based on minimum market size assumptions, distance threshold and population distribution. Alonso’s (1964) additions to location
D
choice modelling emphasize the preference of retailers on locations with many potential customers close by, causing them to outbid other claims for central locations. Hotelling (1929) posed that competing actors often
TE
tend to locate close to each other in most central locations in order to minimize customer loss due to competition and to reach to the largest available market. Thus, competition may be one reason for
EP
agglomeration. Next to competition considerations, diversity benefits (Jacobs, 1969) and scale benefits (Marshall, 1890) are often mentioned as reasons for the occurrence of agglomeration of economic activity (De
CC
Groot et al., 2016). Lastly, it is worth mentioning that the transport cost and agglomeration benefit components of location choice have been tied together in the models of New Economic Geography (Krugman,
A
1996).
3.2. Main assumptions and model parameters As suggested by land-use literature, it may be expected that facilities, public or commercial, look for the best location among available options to satisfy their optimal or at least minimal needs for number of users. They thus require sufficient market (population) potential, i.e., potential users within range, to maximize their profit or to reach the maximum number of people in need of a service. Following Christaller’s theory, services thus 6
select the location that is most central to their intended users, in order to maximize the utility of their location. The method applied in this study is based on the assumption that service location choices only try to maximize market potential. It must be admitted location choice considerations are more complicated in reality, taking into account various additional factors such as transport services, land and rent prices, zonal restrictions and income distribution. Additional assumptions are that the majority of the population should be within acceptable distance of a service, and that each service should cover at least a minimum number of people to
SC RI PT
be economically viable.
Following this approach, this study first identifies the market potentials in Europe using a fine-resolution
population map. Then, it allocates generic facilities based on their potentials at local, subregional and regional levels where they can serve to people within corresponding ‘ideal’ and ‘maximum’ distances. To do so it uses an iterative dynamic discrete spatial allocation procedure. In a subsequent step, the study computes average distance and time per person to the nearest facility as the primary measure of physical access to generic
U
service areas and facilities. The parametrization of market size and geographical catchment area enables the
N
definition of services with different spatial scales, as proposed by Christaller (1933) and Lösch (1938). For
A
example, the majority of users visit local facilities frequently. This makes proximity a key factor for local facilities. Regional facilities, on the other hand, are visited less frequently by the majority of the users. For them
M
the quality or specialization of the service is more important. Therefore, this approach assumes that local services require fewer people in geographically smaller catchment areas, while regional services need more
D
people in a much larger catchment area. This conceptualization has been taken as the base for the allocation of
TE
generic services in this study as illustrated in Figure 1. For example, this approach defines a local facility as serving 5 to 10 thousand people within 5 km distance and
EP
a regional facility as serving 500 thousand to 1 million people within 100 km. For the allocation exercise presented in this paper, generic facilities were allocated with parameters considered characteristics for local, subregional and regional services. Sets of parameters assumed characteristic for the selected service types are
CC
given in Table 1. The values for those criteria were determined with the aim of representing real-life hierarchical structure of different services.
A
A recent typology by European Commission (Dijkstra and Poelman, 2014; EUROSTAT, 2016) for European settlements uses a fine resolution population grid to classify settlements into cities, towns and suburbs, and rural areas. In this method, towns or suburbs should have an urban cluster of at least 5 000 inhabitants, and a city or an urban centre with at least 50 000 inhabitants. To decide on the population thresholds for local and subregional facilities, these two population criteria have been taken as ideal service area populations. The ideal population for regional services has been set at 500 000, which is an increase by another factor of 10. Those 7
population thresholds together with the thresholds for services such as health, public library, market, railway stations and airports are reviewed in Table 2. The criteria for ideal and maximum distance to services have also been determined based on the relevant thresholds used or reported in the literature where available. The ideal distance has been taken as 2.5 km for local, 10 km subregional and 50 km for regional services. As there is too much variation in service area requirements, estimating the exact location of a specific service type is outside the scope of this study. Instead, it proposes a universal method, which can be applied to most types of facilities
SC RI PT
to assess the likely accessibility to services based on a population distribution and transport network.
3.3. Facility allocation method and accessibility computation
The four parameters introduced above are used to distribute the facilities or the service accessibility: a
minimum and an ideal service area population (or market size), and an ideal and a maximum geographic service area size. Given those service-specific parameters, market potentials are computed using a fine
U
resolution population map. Subsequently, the developed method allocates the first facility to the place with
N
the highest market potential within the country of interest. The population served by that facility is subtracted from the available market, and not taken into account during the next market potential computation. The
A
method subsequently searches for the next best location, until all potential locations that meet at least the
M
minimum requirements are exhausted. Thus, when no more locations with ideal population sizes are found, the method searches locations with minimum population sizes, given maximum market potentials. A crucial
D
aspect of the method is that the ideal service area population that the first (best) facility serves, is taken into account in the second iteration, and so on. Table 3 provides a more detailed description of the applied method.
TE
With this algorithm, the services are allocated as point features. Subsequently, a road network dataset together with a population distribution data can be used to compute accessibility to the allocated services. In
EP
the study at hand, Dijkstra shortest-path routines have been used to compute travel distances and travel times to nearest facilities for all populated points in a regularly latticed population map. Those can subsequently be
CC
averaged to administrative units that have political relevance such as the NUTS3 level as shown in Figure 2.
A
3.4. Limitations of the proposed method The proposed method is relatively straightforward, making computation for sizeable territories practical; and it depends mainly on population distribution data with a sufficiently fine resolution. Such data is increasingly available worldwide (European Commission and Columbia University, 2015; GEOSTAT, 2011), and with the increasing availability of methods that can project future population distributions, this method can be used to evaluate the impact of population redistribution trends and related policies on service accessibility. 8
This method, however, is highly simplified. Other potentially important factors are not included. For example, some specific (typically commercial) services may prefer to cluster together in order to benefit from competition or diversity externalities. Some services may have area and ideal population requirements that are more flexible than the ones used here. Others may benefit from economies of scale. The public sector also has both a direct and indirect impact on the distribution of public and private services. Public interventions frequently ensure a higher level of accessibility in more sparsely populated areas. A number of public policies
SC RI PT
will also affect location choices, such zoning regulations or limits on the size of facilities. Furthermore, the
specific structure of public services may vary substantially between countries and affect the spatial distribution of services. A comprehensive EU-wide overview of public interventions affecting location choices or the structure of public services is currently unavailable and well beyond the scope of this method.
Other approaches might take into account car ownership and public transport availability to measure
accessibility. Furthermore, aspects such as the right to use, affordability, quality and acceptability of services
U
(Penchancsky and Thomas, 1981) are crucial elements in accessibility debates, but these are concerns are
N
beyond the scope of this study. This method focuses only on the aspects related to the geographical
A
distribution of services and the performance of road networks. It follows the main assumptions of the central place theory introduced earlier, hence, carry its main shortcomings such as uniform space and homogenous
M
consumer profile. Parr (2017) and Van Meeteren and Poorthuis (2018) can be referred for extensive reviews on this issue.
D
3.5. Selected Data
TE
Generic services have been allocated for all 28 European Union member states. The GEOSTAT (2011) census population grid at 1 km spatial resolution was used to capture population distribution. To measure road
EP
distances and driving times, data from the TELEATLAS MultiNet 2014 road network1 was used. For validation
1
CC
purposes, the data on health facilities from the TOMTOM MultiNet Points of Interest v1.92 database was used.
A
At the time of the computations, the road network for Cyprus was not available. Therefore, its average road distance and average driving time values were interpolated based on the measurements by Euclidean distances. The values for some distant islands / cities such as Melilla, Ceuta and Ajaccio were also estimated based on the Euclidian distances where necessary. 2
Both the TELEATLAS and the TOMTOM datasets are commercial products. Compared to open source alternatives, their main advantage is that they have ready to use information with additional attributes and with less missing data values. The TELEATLAS road network is very detailed and complete in almost all countries and regions of Europe. Its coverage is sufficient for this type of analyses. The TOMTOM points of interests’ coverage for several different services, are exhaustive in some countries and regions, but not sufficient in some others (i.e., missing information of many points of interests).
9
The computations were handled using scripts developed in MATLAB, ARCGIS and GeoDMS. The results were reported by countries, metropolitan regions and by degree of urbanisation3, and at the level of municipalities (local administrative unit level 2 or LAU2) covering all EU28 countries. As road distances and travel times to facilities were found to be highly correlated (e.g., 98% correlation or higher at LAU2 level results), this paper mainly presents distance-based accessibility indicators.
SC RI PT
4. Results 4.1. Mapping accessibility to generic services in Europe
Potential accessibility indicators at a European scale, typically follow a strong core-periphery pattern.
Accessibility is higher in central countries and lower in peripheral countries (see for instance, Spiekermann et al., 2006; 2015; Kompil et al., 2016). This core-periphery pattern differs only slightly if a peripheral region has large population or a high level of economic activity. Unlike the potential accessibility measure, accessibility to
U
generic services have a far more differentiated spatial pattern. Generic services, at least local or everyday
N
services, should be easily accessible for everyone. Therefore, accessibility to local services will vary less than
A
across the EU than subregional or regional services, because they require higher number of users than some remote or rural areas can provide. Access to higher-level services will depend on the spatial distribution of
M
population and regions with similar characteristics should have similar levels of accessibility. The first results of the computations are the grid-based maps that show road (driving) distance to the nearest
D
generic facility across Europe. As indicated in Figure 3, these maps are prepared at one square kilometre spatial
TE
resolution to show distance to the nearest local, subregional or regional facility. Figure 4 shows aggregated average results of those grid maps by municipalities (LAU2) as a measure of generic accessibility at local,
EP
subregional and regional levels. Finally, Figure 5 gives modelled country and EU averages for accessibility to services at local, subregional and regional levels.
CC
As expected, all of these figures indicate that local services have better accessibility compared to the subregional and regional services and it is more homogenously distributed across Europe. It is equally high
A
almost everywhere, especially in urbanized areas, and varies slightly in rural areas. On the other hand, people
3
Metropolitan regions approximate Functional Urban Areas (FUA - cities and their commuting zones), with at least 250 thousand inhabitants, as covered by one or more NUTS3 regions. The degree of urbanization is a classification of local administrative units (LAUs) that indicates the characteristics of a particular area, based on a population grid composed of 1 km² cells (and clusters thereof), identifying: cities (densely populated areas), towns and suburbs (intermediate density areas) and rural areas (thinly populated areas). For further information on those geographical units see EUROSTAT (2016) and Dijkstra and Poelman (2014).
10
are expected to travel further to reach subregional and regional services and the travel range is even higher in rural or remote rural areas. Based on the modelled results, in order to reach to the nearest local facility people are expected to drive, on average, 4 km for local services, 9 km for subregional services, 30 km for regional services, which would take approximately 9, 14 and 30 minutes respectively. At the country level, Malta, the Netherlands, Belgium, the UK and Germany are expected to have a high level of
SC RI PT
service accessibility. Thanks to their highly urban and concentrated population, they have the lowest average distance values in Europe at almost all levels. For most other countries, average distances are much higher for regional services; although they perform similarly in accessibility to local and subregional services. According to the simulation results, residents of countries with particularly sparse population distributions, such as Finland, Sweden, Latvia and Estonia, are expected to face with longer distances to reach regional services (Figure
5).Within countries, average distance to different services in the metropolitan regions are much lower than in non-metropolitan regions (Figure 6 and Figure 7). Overall, average distance to the nearest local facility is
U
estimated as 3.2 km in metropolitan and 6.7 km in non-metropolitan regions. The gap is even wider for
N
regional services, i.e. 21 km in metropolitan and 61 in non-metropolitan regions. The capital metropolitan
A
regions are among the best performing regions in the majority of countries.
M
Figure 8 and Figure 9 show accessibility levels by degree of urbanisation and the pattern is clear: the higher the level of urbanisation, the better the level of accessibility to facilities. Cities provide higher level of accessibility to all type of service areas with only small differences between countries. An average person living in a city
D
would have a local facility within 2.5 km, while for someone living in a rural area it is 9 km away. The distance
TE
to local services in rural areas differs more between countries, from 2 kilometres in Malta to 21 kilometres in Finland. The gap between cities and rural areas and the variation between rural areas are even bigger when
EP
higher-level services are considered. For instance, the average distance per person to the nearest facility in cities is estimated as 2 km for local, 4 km for subregional and 12 km for regional facilities. It is much higher in
CC
rural areas with 8 km for local, 18 km for subregional and 48 km for regional facilities. The level of accessibility to services not only changes by urban-rural typology, but also by population size of the city. Larger cities perform better in particular with regard to accessing regional services (Figure 10), which may
A
be considered a logical consequence of the minimum market size criterion in the outlined method. A person living in a city with less than 100 thousand inhabitants is expected to travel in average 30 km to reach a (generic) regional facility, whereas one can find a regional facility within 6-8 kilometres in cities with more than 1 million inhabitants. For local and subregional services, however, there is no additional benefit linked to city size. All cities, including the small ones have a local service within 2 kilometres and a subregional service within 4 km. 11
Instead of looking at the average distance, this method can also measure the share of people living within a certain distance. This is useful as it can assess the potential for walking or cycling to accessibility these services. Figure 11 shows the share of population that has a facility within a certain road distance for the three types of services. The simulation results indicate that 81% of the people in Europe are expected to have a local service within 5km distance. It is 49% for subregional and 15% for regional services. If you go further to 15 kilometres, 96% of the population would have a local, 85% has a subregional and 41% has regional services accessible to
SC RI PT
them. A more pessimistic reading of the simulation graphs could highlight that one out of four Europeans are expected not to have a generic regional facility within 40 kilometres. Finally, as indicated in Figure 12, large cities offer the possibility of accessing services by walking or by bicycle while in rural areas or in smaller towns, it is much more difficult to do so. For instance, the average share of population in the EU living within 1 km of local services is estimated to increase rapidly with the degree of urbanisation and the size of city, rising from 12% in rural areas to over 80% in cities of more than 5 million inhabitants. These results show that cities have a
U
greater potential to access services by walking and cycling. To realise such a potential, this does require that
A
4.2. Comparison and validation of the simulated results
N
the appropriate infrastructure is in place that the traffic safety is not deterrent to using active modes.
M
To validate the accuracy of the proposed method, the simulated results on accessibility to generic services have been compared with observed results on accessibility to health services. To do so, data on health facilities from the TOMTOM MultiNet Points of Interest v1.9 database was used. That dataset includes location information
D
on health services for all European countries. However, the available information for health services is not
TE
completely representative for all countries; a quick comparison of observed health facilities per capita indicates that there are likely considerable omission errors in some of the countries. In the selected eight countries
EP
(Belgium, Estonia, France, Greece, Latvia, Lithuania, the Netherlands and the United Kingdom) 7 to 24 facilities per 100 000 inhabitants are observed, while other European countries average 1.7 health facility per 100 000
CC
inhabitants. It has to be emphasized that the higher incidence of health facilities in the selected countries does not imply there are no data omissions in those countries, but at least the omissions are comparably much smaller. For road distances to the nearest health facility (at any level), again the data from the TELEATLAS
A
MultiNet (2014) road network was used. The data on observed health facilities include both typically local services (e.g. health care services and polyclinics) and larger-scale services (e.g. university hospitals) without any differentiation. Therefore, the simulation results were also computed using all type of facilities (local, subregional and regional) without any differentiation and measured as average road distance per person to the nearest generic facility. Both observed and simulated accessibility results are mapped in Figure 13.
12
In addition to this, comparison of observed and simulated results on accessibility are given in Table 4 together with selected goodness-of-fit statistics. In the selected eight countries, based on 1km grid aggregation, average distance per person to the nearest observed health facility is computed as 4.3 kilometres. On aggregate, the proposed method underestimated observed distances with 700 metres. This result points out a tendency that the developed approach is estimating less distance than the reality, at least for health services. It is something that might be expected, if the following two facts are taken into consideration: (I) Generic services are
SC RI PT
universal and could be taken as any type of service. Some services might be found more common across space than the health services, like schools or pharmacies. Therefore, to have some small distortions from the reality is inevitable and should be accepted, if the overall accessibility pattern is not affected at all. (II) Another reason might be that the data could be missing some of the health services in some of the countries, which increase average distance per person to the nearest health facility. However, since the amount of the difference caused
A
CC
EP
TE
D
M
A
N
U
by those missing facilities cannot be estimated, it is treated as a complete dataset for the selected countries.
13
N U SC RI PT
The difference of 700 meters shows 16% deviation from the observations. For some countries like the Netherlands (2.2 km observed and 2.3 km simulated), Belgium (3.1 km observed and 2.8 km simulated) and France (4.6 km observed and 4.3 km simulated), the deviation is even smaller, which lies between 4% and 9%. On the other hand, the United Kingdom (4.1 km observed and 2.7 km simulated) deviates 34% and Lithuania (4.8 km observed and 6.4 km simulated) deviates 33% from the observed average distances to health facilities. Although omission errors in the observed data cannot be ruled out, it is most likely that processes unobserved in the proposed allocation method are causing the found discrepancies between observed and
A
simulated outcomes. For France (Figure 13) it is immediately clear that facilities are reasonably evenly spread out over its territory and located even where population numbers may not warrant a health facility. The discrepancies are not caused by excessive allocation of facilities, as we observe 13,
M
and allocate 12 facilities per 100,000 inhabitants. A likely explanation of the found differences may be sought in public intervention, which could have led to seemingly underserved populous areas, causing higher real distances per capita then the allocated facilities would yield. It must be emphasized
ED
that the capacity and quality of health facilities is not taken into account, so that seeming underserving is solely a question of geographical accessibility. In any case, public intervention may be an important factor to explain discrepancies in the allocated results. For example, Lithuania has many more health facilities than its population size would warrant and they are distributed much more frequently and homogenously than its population
PT
distribution would suggest. In this case, public intervention may have led to overserved areas or lower average per-capita distance to health services
CC E
than the proposed method yields.
The second part of Table 4 includes comparisons established at the level of communes (LAU2), the smallest zonal aggregation used in the study. Since the accessibility results are (designed to be) reported based on some zonal aggregations, it is more significant to elaborate this stage of comparison with some selected goodness-of-fit statistics. These eight EU countries have over 48.000 municipalities with 9 km overall average distance per person
A
to the nearest health facility. The simulated average distance on the other hand is 9.8 km, which points out an 800 meters and 9% deviation from the observations. The Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) for this LAU2 level comparison are 3.9 km and 54% respectively. In terms of these two statistics, the developed approach best performs in the Netherlands, Belgium and Latvia. Share of LAU2 zones with a difference less than half of the Average Distance (SEAD) is 73%. This is a very important and promising result and means that almost ¾ of LAU2 values for accessibility to services are estimated with very small discrepancies. This rate is even much higher (up to 95%) in the above-mentioned three countries. Finally, the correlation (measured with Pearson’s r correlation coefficient) between the observed and simulated values is also quite high with 14
N U SC RI PT
an overall rate of 65%. It is a reasonably successful rate if the fact that the estimations are based only on one variable (population) and some rough assumptions on the market requirements of facility types.
Comparing and visualizing the overall distribution / pattern of accessibility to services is as important as mean absolute error and correlation coefficient measurements. Figure 14 shows average road distance per person to the nearest health (observed) and generic (simulated) facility by cumulative share of population for eight countries and Belgium. According to this, accessibility to services both observed and simulated follow a power-law distribution,
A
the smaller the distance to local services, the higher the share of population. Overall, the share of population with a health facility within 5 km is 75%, while it is simulated with generic services as 84%. This result is quite satisfactory in terms of capturing the main pattern in accessibility to services.
M
However, the main tendency to overestimate the share of the population with smaller average distances due to the reasons remarked before is more visible with this analysis. Figure 14 also shows the comparison results for Belgium, which has been selected because of its extensive data coverage on
ED
health services and urban-rural structure. The simulated share of population with corresponding average distance fit better to the observed ones in Belgium. For instance, the share of population with a health facility within 3 km is 59%, while it is simulated with generic facilities as 65%.
PT
It is worth monitoring also spatial distribution of accessibility to services and compare observed versus simulated results on space visually. Figure 15 is prepared for this purpose and compares the observed values (above) with the simulated ones (below) at the level of 1 km grids and communes (LAU2).
CC E
An overall evaluation of those maps indicate that the developed approach is quite successful in capturing the real-life pattern of accessibility to services in Belgium. It has identified the majority of the approximate facilities in the main population centres and sub-centres (coloured in red and dark orange) successfully; and estimated accessibility to services with a very high accuracy in Belgium. The main regions that the developed approach cannot fully
A
capture the accessibility pattern are became some of the remote and /or rural regions (coloured in dark green) in the south of the country.
4.3. Summary of the findings The modelled results show that local services such as schools, small health facilities and small shops are generally expected to be easily accessible in all type of municipalities in Europe, though they take longer to reach in suburban and especially rural areas. The average distance is expected to be greater for subregional services such as high schools, hospitals, theatres and supermarkets, and greatest of all for regional services, such as specialized centres 15
N U SC RI PT
for education and health, large facilities for sports and cultural activities and government offices. Based on the simulation results, one would have to drive 4 km to reach a local service, 9 km for a subregional, and 30 kilometres for a regional service.
Cities provide better opportunities in accessing services. As suggested by the modelled results, average distance per person to the nearest facility in cities is estimated as 2 km for local, 4 km for subregional and 12 km for regional facilities. It is much higher in rural areas with 8 km for local, 18 km for subregional and 48 km for regional facilities. Metropolitan regions perform almost three times better than non-metropolitan regions in terms of
A
accessibility to generic services. Capital metropolitan regions are expected to be among the best performing regions in the majority of countries in terms of accessibility to services.
M
It has been estimated that four out of five people in Europe have at least one local or daily service within 5 kilometres. On the other hand, in Europe, one out of four people lack a regional or high-order service within 40 kilometres. Finally, cities offer far more potential to access services by walking
ED
and cycling. Based on the model, the population share within 1 km of a local service rises from 12% in rural areas, to 35% in towns and suburbs, 50% in small cities and to over 80% in the largest cities.
PT
The proposed model is highly capable of capturing real life patterns in accessibility to services. The results of this model correspond closely to the spatial distribution of observed health care services, as shown in the comparison and validation section. It offers a promising line for further research to
CC E
understand the possibilities and difficulties to improve accessibility to services in different types of territories. Using only a few different types of input data (a population grid and a road network), it produces results in line with previous works by Milbert et al. (2013), Rauhut and Komornicki (2015), Papaioannou and Wagner (2017) and ESPON (2017a; 2017b). In particular, this model captures well the gap between urban and rural areas, and the
A
service hierarchy in terms of accessibility.
5. Conclusion Recent progress in availability of more detailed spatial data and increased computational capacity give researchers the opportunity to explore new aspects of accessibility. The combination of detailed data on the spatial distribution of population, activities and facilities, and transport means and services, lately, made it possible to explore different characteristics of accessibility to service areas and facilities. At this point, this study developed a 16
N U SC RI PT
novel approach to simulate (potential) location of facilities where such information is missing; proposed a universal method based on a detailed population distribution to measure accessibility to generic services; and finally, analysed the characteristics of European cities and regions in terms of modelled accessibility to service areas and facilities. The outcome of this analysis provides useful insights and sound evidence to improve knowledge on conditions of accessibility to services across the EU.
The two main advantages of this study are that it produces a pan European data set with the location proxies of services and that it provides a clear
A
benchmark against which current and future services can be assessed in terms of accessibility. Since it removes data availability problems, the model is very suitable for land use and transport planning as it can test the impact of alternative scenarios of population and infrastructure development on the
M
number of service locations and trip lengths needed to reach them. In addition to this, it can be used for comparing real landscapes with idealized landscapes to highlight overserved and underserved areas by a specific service. A successive study could explore, for example, the relative benefits of
ED
a) adding more service points versus improving the road network and b) encouraging less dispersed settlements, or the impacts of c) emerging transport technologies or changing demographic dynamics, e.g., autonomous cars and aging population. This model could also be used to estimate the
PT
potential accessibility benefits of cross border services.
A
CC E
Disclaimer: The views expressed are purely those of the authors and may not in any circumstances be regarded as stating an official position of the European Commission.
17
N U SC RI PT
References
Aderamo, A. J. and Aina O. A. (2011), Spatial inequalities in accessibility to social amenities in developing countries: a case from Nigeria, Australian Journal of Basic and Applied Sciences, 5(6), 316-322. Alonso, W., 1964, Location and land use: toward a general theory of land rent, Cambridge: Harvard University Press. Banister, D., 2008, The sustainable mobility paradigm, Transport Policy, Volume 15 (2), 73-80, https://doi.org/10.1016/j.tranpol.2007.10.005. Bertolini, L., 2005, Sustainable urban mobility, an evolutionary approach, European Spatial Research Policy, 12 (1), 109-126.
M
A
Bibri, S.E. and Krogstie, J., 2017, Smart sustainable cities of the future: An extensive interdisciplinary literature review, Sustainable Cities and Society, Vol. 31, 183-212, https://doi.org/10.1016/j.scs.2017.02.016. Bourdic, L., Salat S. and Nowacki C., 2012, Assessing cities: a new system of cross-scale spatial indicators, Building Research & Information, 40:5, 592605, https://doi.org/10.1080/09613218.2012.703488.
ED
Breuer, I.M., Milbert, A., Foss O., Humer A., Palma P., Rosik P., Stepniak M. and Velasco X., 2013, European Atlas of Services of General Interest, as an output of the ESPON project entitled ‘Indicators and Perspectives for Services of General Interest in Territorial Cohesion and Development’.
PT
Caubel, D., 2006, An Increase Of Public Transport And Accessibility To Urban Amenities, Some Limited Results: The Case Of The Lyons Conurbation, WIT Transactions on The Built Environment, Vol 89, 507-516, doi:10.2495/UT060501 Christaller, W., 1933, Die zentralen Orte in Süddeutschland, Gustav Fischer, Jena.
CC E
Curtis C. and Scheurer J., 2010, ‘Planning for sustainable accessibility: developing tools to aid discussion and decision-making, Progress in Planning, 74 (2), 53–106. https://doi.org/10.1016/j.progress.2010.05.001 Curtis, C., 2008, Planning for sustainable accessibility: The implementation challenge, Transport Policy, 15 (2), 104-112, https://doi.org/10.1016/j.tranpol.2007.10.003.
A
de Groot, H. L. F., Poot, J. and Smit, M. J., 2016, Which agglomeration externalities matter most and why? Journal of Economic Surveys, 304, 756–782. http://doi.org/10.1111/joes.12112 Dijkstra, L. and Poelman, H., 2014, A harmonised definition of cities and rural areas: the new degree of urbanization, Regional Working Papers, WP 01/2014. E, http://ec.europa.eu/regional_policy/sources/docgener/work/2014_01_new_urban.pdf - access 01/09/2017. Donnelly, F. P., 2015, Regional variations in average distance to public libraries in the United States, Library & Information Science Research, Volume 37 (4), 280-289, https://doi.org/10.1016/j.lisr.2015.11.008. EC, 2018a, The urban agenda for the EU, http://ec.europa.eu/regional_policy/en/policy/themes/urban-development/agenda/ - access 01/09/2018. EC, 2018b, Urban mobility: Strategies and policies, https://ec.europa.eu/info/node/3859 - access 01/09/2018. 18
N U SC RI PT
ESPON, 2017a, PROFECY – Processes, Features and Cycles of Inner Peripheries in Europe (Inner Peripheries: National territories facing challenges of access to basic services of general interest), Final Report, Version 07/12/2017. ESPON, 2017b, PROFECY – Processes, Features and Cycles of Inner Peripheries in Europe (Inner Peripheries: National territories facing challenges of access to basic services of general interest), Annex 2. Datasets and Database, Final Report, Version 07/12/2017. European Commission, 2004, A new partnership for cohesion: convergence competitiveness cooperation, Third report on economic and social cohesion, Luxembourg: Publications Office of the European Union, ISBN 92-894-4911-X.
A
European Commission, Joint Research Centre (JRC); Columbia University, Center for International Earth Science Information Network - CIESIN (2015): GHS population grid, derived from GPW4, multitemporal (1975, 1990, 2000, 2015). European Commission, Joint Research Centre (JRC) [Dataset] PID: http://data.europa.eu/89h/jrc-ghsl-ghs_pop_gpw4_globe_r2015a
M
EUROSTAT, 2016, Urban Europe — statistics on cities, towns and suburbs, Publications office of the European Union, Luxembourg, ISBN 978-92-7960139-2, DOI: 10.2785/91120.
ED
Fassmann, H., Rauhut, D., da Costa E.M. and Humer, A., 2015, Services of general interest and territorial cohesion: European perspectives and national insights. V&R Unipress GmbH, Göttingen, Germany. ISBN 978-3-8471-0471-1. Ferreira, A., Beukers, E. and Te Brömmelstroet, M., 2012, Accessibility is Gold, Mobility is Not: A Proposal for the Improvement of Dutch TransportRelated Cost-Benefit Analysis, Environment and Planning B: Planning and Design, vol. 39, 683 – 697, https://doi.org/10.1068/b38073.
PT
Garretsen, H. and Marlet G., 2017, Amenities and the attraction of Dutch cities, Regional Studies, 51:5, 724–736, http://dx.doi.org/10.1080/00343404.2015.1135239
CC E
GEOSTAT, 2011, GEOSTAT 1 km2 population grid, http://ec.europa.eu/eurostat/web/gisco/geodata/reference-data/population-distributiondemography/geostat - access 01/09/2017. Geurs, K. T., and van Wee, B., 2004, Accessibility evaluation of land-use and transport strategies: review and research directions, Journal of Transport Geography, 122, 127–140. https://doi.org/10.1016/j.jtrangeo.2003.10.005
A
Ghosh-Dastidar, B., Cohen, D., Hunter, G., Zenk, S. N., Huang, C., Beckman, R. and Dubowitz, T., 2014, Distance to Store, Food Prices, and Obesity in Urban Food Deserts, American Journal of Preventive Medicine, 475, 587–595. https://doi.org/10.1016/j.amepre.2014.07.005 Gutiérrez J., 2009, Transport and Accessibility, In Kitchin, R. and Thrift, N. (Eds), International Encyclopedia of Human Geography, pp. 410-417, Oxford: Elsevier. Hirai, H., Kondo, N., Sasaki, R., Iwamuro, S., Masuno, K., Ohtsuka, R., Miura, H. and Sakata, K., 2015. Distance to retail stores and risk of being homebound among older adults in a city severely affected by the 2011 Great East Japan Earthquake, Age and Ageing, 443, 478–484. http://doi.org/10.1093/ageing/afu146 Hotelling, H., 1929, Stability in competition, The Economic Journal, 39153, 41–57. 19
N U SC RI PT
Jacobs, J., 1969, The economy of cities, Vintage Books, New York.
Jacobs-Crisioni, C., Batista e Silva, F., Lavalle, C., Baranzelli, C., Barbosa, A. and Castillo, C. P., 2016, Accessibility and territorial cohesion in a case of transport infrastructure improvements with changing population distributions, European Transport Research Review, 89. http://doi.org/10.1007/s12544-016-0197-5 Keeble, D., Owens, P. L. and Thompson, C., 1982, Regional accessibility and economic potential in the European Community, Regional Studies, 166, 419–432. http://dx.doi.org/10.1080/09595238200185421
A
Kompil, M., Demirel, H. and Christidis, P., 2016, Accessibility and territorial cohesion: ex post analysis of Cohesion Fund infrastructure projects, in Geurs, K. T., Patuelli, R., and Dentinho, T. P. (eds) Accessibility, Equity and Efficiency. Cheltenham, the UK, Edward Elgar Publishing, 210–242. http://dx.doi.org/10.4337/9781784717896.00021
M
Krugman, P., 1996, Urban concentration: the role of increasing returns and transport costs, International Regional Science Review, 19, 5–30. https://doi.org/10.1177/016001769601900202
ED
López, E., Gutiérrez, J. and Gómez, G., 2008, Measuring regional cohesion effects of large-scale transport infrastructure investments: An accessibility approach, European Planning Studies, 162, 277–301. http://dx.doi.org/10.1080/09654310701814629 Lösch, A., 1938, The Nature of Economic Regions. Southern Economic Journal 5:71–78.
PT
Lösch, A., 1954, The Economics of Location, Translated from the second revised edition by W. H. Woglom with the assistance of W. F. Stolper, New Haven and London: Yale University Press. Originally published in German as Die räumliche Ordnung der Wirtschaft first in 1940, and in a revised second edition in 1944.
CC E
Lucas, K., 2012, Transport and social exclusion: Where are we now?, Transport Policy, 20, 105–113. http://doi.org/10.1016/J.TRANPOL.2012.01.013 Lucas, K., van Wee, B. and Maat, K., 2016, A method to evaluate equitable accessibility: combining ethical theories and accessibility-based approaches, Transportation, 433, 473–490. http://doi.org/10.1007/s11116-015-9585-2
A
Marques da Costa, E., Costa, Palma, P. and Marques da Costa, N., 2015, Regional Disparities of SGI provision, In Fassmann, H., Rauhut, D., da Costa, E.M. and Humer, A. (Eds.), Services of general interest and territorial cohesion: European perspectives and national Insights. pp. 92-121. V&R Unipress GmbH, Göttingen, Germany. ISBN 978-3-8471-0471-1. Marshall, A., 1890, Principles of economics: An introductory volume, Macmillan, London. McDonald, N. C., 2008, Children’s mode choice for the school trip: the role of distance and school location in walking to school, Transportation, 35, 23– 35. https://doi.org/10.1007/s11116-007-9135-7 McGrail, M. R., 2012, Spatial accessibility of primary health care utilising the two step floating catchment area method: an assessment of recent improvements, International Journal of Health Geographics, 11:50. https://doi.org/10.1186/1476-072X-11-50
20
N U SC RI PT
Milbert, A., Breuer, I.M., Rosik, P., Stepniak, M. and Velasco, X., 2013, Accessibility of services of general interest in Europe, Romanian Journal of Regional Science, Special Issue on Services of General Interest, Vol. 7, 37–65. http://www.rrsa.ro/rjrs/V7SP1.Milbert.pdf Nicholl, J., West, J., Goodacre, S. and Turner, J., 2007, The relationship between distance to hospital and patient mortality in emergencies: an observational study, Emergency Medicine Journal, 249, 665–668. http://doi.org/10.1136/emj.2007.047654 Öner, Ö., 2017, Retail city: the relationship between place attractiveness and accessibility to shops, Spatial Economic Analysis, 121, 72–91. https://doi.org/10.1080/17421772.2017.1265663
A
Páez A., Scott D.M. and Morency, C., 2012, Measuring accessibility: positive and normative implementations of various accessibility indicators, Journal of Transport Geography, 250, 141–153. https://doi.org/10.1016/j.jtrangeo.2012.03.016
M
Pakoz M. Z. and Yuzer M. A., 2014, Determinants of access to healthcare: A survey in Istanbul, 54th Congress of the European Regional Science Association: "Regional development & globalisation: Best practices", 26-29 August 2014, St. Petersburg, Russia, https://www.econstor.eu/handle/10419/124488
ED
Papaioannou, D. and Wagner. N., 2018, Measuring access to healthcare and education by car and public transport in 18 cities across the world, TRB 97th Annual Meeting, January 7-11, Washington D. C. Parr, J. B., 2017, Central place theory: An Evaluation. Review of Urban & Regional Development Studies, 11(3), 256–14.
PT
Penchansky, R. and Thomas, J.W., 1981, The concept of access: definition and relationship to consumer satisfaction, Medical care, 19. 127-40. http://dx.doi.org/10.1097/00005650-198102000-00001
CC E
Pereira R. H. M., Schwanen T. and Banister D., 2017, Distributive justice and equity in transportation, Transport Reviews, 37:2, 170-191, https://doi.org/10.1080/01441647.2016.1257660 Pilkington, H., Prunet, C., Blondel, B., Charreire, H., Combier, E., Le Vaillant, M., Amat-Roze J.M. and Zeitlin, J., 2017, Travel time to hospital for childbirth: comparing calculated versus reported travel times in France, Maternal and Child Health Journal. http://doi.org/10.1007/s10995-0172359-z
A
Rauhut, D. and Komornicki, T., 2015, The challenge of SGI provision in rural areas, 55th Congress of the European Regional Science Association: ‘World Renaissance: Changing roles for people and places’, 25-28 August 2015, Lisbon, Portugal. hdl.handle.net/10419/124605 Rauhut, D., Smith, C., Humer, A., Ludlow, D. and Borges, L., 2013, Indicators and perspectives for Services of General Interest in Territorial Cohesion and Development (SeGI). Final Report | Version 25/05/2013. ESPON & Royal Institute of Technology KTH. Rietveld P. and Bruinsma F., 1998, Is Transport Infrastructure Effective? Transport Infrastructure and Accessibility: Impacts on the Space Economy, Springer-Verlag, Berlin.
21
N U SC RI PT
Spiekermann K., Wegener M., Kveton V., Marada M., Schürmann C., Biosca O., Ulied Segui A., Antikainen H., Kotavaara O., Rusanen J., Bielanska D., Fiorello D., Komornicki T. and Rosik P., 2015, ‘TRACC: Transport Accessibility at Regional/Local Scale and Patterns in Europe’, Final Report | Version 06/02/2015, Volume 2 - TRACC scientific report, ESPON & Spiekermann & Wegener, Urban and Regional Research S&W. Spiekermann K., Wegener M., Kveton V., Marada M., Schürmann C., Biosca O., Ulied Segui A., Antikainen H., Kotavaara O., Rusanen J., Bielanska D., Fiorello D., Komornicki T. and Rosik P., 2015, TRACC: Transport Accessibility at Regional/Local Scale and Patterns in Europe, Final Report | Version 06/02/2015, Volume 2 - TRACC scientific report, ESPON & Spiekermann & Wegener, Urban and Regional Research S&W. Spiekermann, K. and Wegener M., 2006, Accessibility and spatial development in Europe, Scienze Regionali, 5 (2), 15–46.
A
Stepniak, M. and Rosik, P., 2013, Accessibility improvement, territorial cohesion and spillovers: a multidimensional evaluation of two motorway sections in Poland, Journal of Transport Geography, 31, 154–163. http://doi.org/10.1016/j.jtrangeo.2013.06.017
M
UN, 2018, Goal 11: Sustainable cities and communities, http://www.undp.org/content/undp/en/home/sustainable-development-goals/goal-11sustainable-cities-and-communities.html#targets - access 01/09/2018.
ED
Van Meeteren, M., and Poorthuis, A., 2018, Christaller and “big data”: recalibrating central place theory via the geoweb. Urban Geography, 39(1), 122– 148 van Wee, B. and Geurs, K., 2011, Discussing equity and social exclusion in accessibility evaluations, European Journal of Transport and Infrastructure Research, 114, 350–367. http://www.ejtir.tudelft.nl/issues/2011_04/abstracts/2011_04_01.html
PT
Vickerman, R., Spiekermann, K. and Wegener, M., 1999, Accessibility and economic development in Europe. Regional Studies, 331, 1–15. http://dx.doi.org/10.1080/00343409950118878
CC E
von Thünen, J. H., 1826, Der isolierte staat in beziehung auf landwirtschaft und nationalökonomie, Hamburg, Perthes. English translation by C.M. Wartenberg: The isolated state, Oxford, Pergammon Press, 1966.
A
Wegener M, Eskelinnen H, Fürst F, Schürmann C, and Spiekermann K., 2001, Criteria for the Sspatial differentiation of the EU territory: geographical position, Forschungen 102.2, Bundesamt für Bauwesen und Raumordnung, Bonn.
22
N U SC RI PT
M
A
Figure 1: Some conceptual assumptions used for the allocation of generic services
A
CC E
PT
ED
Figure 2: Access to generic services – the main computational steps illustrated in Belgium as an example
23
N U SC RI PT A M ED PT CC E A Figure 3: Distance to the nearest facility, at local, subregional and regional levels, by 1 km grids
24
N U SC RI PT A M ED PT CC E
A
Figure 3a: Road distance to the nearest (generic) local facility serving 5 to 10 thousand people, in kilometres (schools, small health facilities, childcare services, sport facilities, small markets etc.)
Figure 3b: Road distance to the nearest (generic) subregional facility serving 50 to 100 thousand people, in kilometres (high schools, hospitals, theatres, cultural facilities, supermarkets, hobby markets etc.)
Figure 3c: Road distance to the nearest (generic) regional facility serving 500 thousand to 1 million people, in kilometres (Specialized centers for education and health, large sportive and cultural centers, governmental organizations, high-tech services etc.)
Figure 4: Accessibility to the nearest facility, at local, subregional and regional levels, by communes 25
N U SC RI PT A M ED PT CC E
A
Figure 4a: Average (road) distance per person to the nearest (generic) local facility serving 5 to 10 thousand people, in kilometres
Figure 4b: Average (road) distance per person to the nearest (generic) subregional facility serving 50 to 100 thousand people, in kilometres
Figure 4c: Average (road) distance per person to the nearest (generic) regional facility serving 500 thousan to 1 million people, in kilometres
Figure 5: Accessibility to the nearest facility, at local, subregional and regional levels, by countries
26
N U SC RI PT A M ED PT CC E A
Figure 6: Accessibility to the nearest local facility by metropolitan regions
27
N U SC RI PT A M ED PT CC E
A
Figure 7: Accessibility to the nearest regional facility by metropolitan regions
28
N U SC RI PT A M ED PT CC E
A
Figure 8: Accessibility to the nearest local facility by degree of urbanisation
29
N U SC RI PT A M ED PT CC E
A
Figure 9: Accessibility to the nearest regional facility by degree of urbanisation
30
N U SC RI PT A M ED PT CC E
A
Figure 10: Accessibility to the nearest facility by city size and degree of urbanisation
31
N U SC RI PT A M ED PT CC E
A
Figure 11: Accessibility to the nearest facility by cumulative share of population
32
N U SC RI PT A M ED PT CC E
A
Figure 12: Share of population living within 1 km distance of a facility by city size and degree of urbanisation
33
34
A ED
PT
CC E A
M
N U SC RI PT
N U SC RI PT
A
CC E
PT
ED
M
A
Figure 13: Distance to the nearest facility, by 1 km grids – observed health facilities vs. simulated (generic) facilities
Figure 13a: Observation - road distance to the nearest health facility, in kilometres, 2014
Figure 13b: Simulation - road distance to the nearest (generic) facility serving 5 thousand to 1 million people, in kilometres
35
N U SC RI PT
A
CC E
PT
ED
M
A
Figure 14: Distance to the nearest facility by share of population – observed versus simulated results
Figure 15: Access to the nearest facility – observed versus simulated results in Belgium
36
37
A ED
PT
CC E A
M
N U SC RI PT
Table 1: Type of services with the corresponding population and distance criteria
Type of services areas / facilities
Ideal service area population
Ideal service area distance
Minimum service area population
Maximum service area distance
10.000 people
2.5 km
5.000 people
5 km
100.000 people
10 km
50.000 people
25 km
1.000.000 people
50 km
500.000 people
100 km
(Schools, small health facilities, childcare services, sport facilities, small markets etc.)
SC RI PT
Local (neighborhood) facilities
Subregional (municipal) facilities (High schools, hospitals, theatres, cultural facilities, supermarkets, hobby markets etc.)
A
CC
EP
TE
D
M
A
N
(Specialized centers for education and health, large facilities for sports and cultural activities, governmental organizations, other high-tech services etc.)
U
Regional facilities
38
Table 2: A review of studies with relevant population and distance criteria
Reference
Case study
Type of measurement
Service type
Caubel (2016)
Lyon, France
Travel time by car
Basket of goods
Local
---
Market
Local
10.000 – 25.000
Health services
Local
2.500 – 5.000
2 km – 4.5 km 5 km
Milbert et al. (2013)
Local
---
Istanbul, Turkey
Travel time (multimodal)
Hospital
Subregional
---
USA
Road distance
Public libraries
Local
---
Primary schools and pharmacies
Local
Europe (five regions from AT, DE, HU,ES,PL))
Railway stations Travel time by car
Hospitals
Victoria, Australia
Travel time by car
A
CC
McGrail (2012)
EP
TE
Airports
EUROSTAT (2016)
Europe
---
Health services
Urban cluster – towns and suburbs High-density cluster - cities Metropolitan regions
Additional notes
10 minutes
Average distance to shops, health and administrative services.
3.5 km – 6 km
SC RI PT
Health services
N
U
Road distance
Distance criterion
30 minutes 3.2 km (2 miles)
---
8 minutes
A
Donnelly (2015)
Population criterion
European cities
---
15 minutes
Subregional
---
10 – 30 minutes
Regional
---
20 - 60 - 130 minutes
Local
5.000 – 25.000
Subregional
25.000100.000
Regional
> 100000
Local
5.000 and higher
Subregional
50.000 and higher
Subregional
M
Dimitrios and Wagner (2017) Pakoz and Yuzer (2014)
Nigeria
Walking radius
D
Aderamo and Aina (2011)
Service scale
Regional
250.000 and higher
10 - 60 minutes (Minimum and maximum catchment size)
Actual population and distance interval from various districts. 80% of the population live within 5 km and less. Acceptable distance for 90% of the surveyed population. 65% of the population live within 3.2 kilometres and less. 80% of the population live within 8 minutes and less. 60% of the population live within 15 minutes and less 10 – 30 minutes (for the majority) 20 to 60 minutes population weighted average in selected regions of ES, DE, AT, PL and 130 minutes in Hungary. Very small rural to medium rural communities. Large rural community. Metropolitan area. Key methodological concepts for EU statistics on territorial typologies.
---
Minimum population requirement other than additional density criteria applicable.
39
Table 3: The algorithm on facility allocation and accessibility computation
2) 3) 4)
Start with a population raster (e.g., GEOSTAT (2011) population grid at 1 km level or lower). Take a zone, in this case a country, and search for its population surface cell by cell (grid by grid) as described in the subsequent steps. Decide on the type of facility (e.g., local or subregional) and apply all corresponding criteria along the rest of the steps (see Table 1 for relevant criteria). The first facility should be located to the cell / place with the highest population (market) potential within the zone (country), based on the ideal distance criteria (e.g., 2.5 km for local and 10 km for subregional facilities). The market (population) potential of a grid cell (𝑀𝑃𝑖𝑖𝑑𝑒𝑎𝑙 ) can be computed using the sum of the populations (𝑃𝑖 ) weighted by the distances (𝐷𝑖𝑗 ) as shown below. The cells out of the ideal distance are not taken into consideration during this step of the calculations.
SC RI PT
1)
𝑀𝑃𝑖𝑖𝑑𝑒𝑎𝑙 = ∑𝑗 𝑃𝑖𝑗 𝐷𝑖𝑗 −1 𝑤ℎ𝑒𝑟𝑒 𝐷𝑖𝑗 ≤ 𝑡ℎ𝑒 𝑖𝑑𝑒𝑎𝑙 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑓𝑎𝑐𝑖𝑙𝑖𝑡𝑦
The ideal service area population (e.g. 10.000 or 100.000) that the first (best) facility serve should not be taken into account in the second iteration. It should be subtracted from the starting matrix or raster based on their distances to the selected location. Starting from the cell where a facility is allocated, the closest cells’ population should be subtracted first; this should continue until the sum of subtractions reach to the level of ideal service area population. For instance, the search could follow a counter clockwise pattern as shown in following illustration. For instance, subtract population from cell (3,3) then cell (3,2) and cell (4,3) and so on so forth. When it exceeds ideal population such as 10.000, return the extra amount of population to the last cell processed. The second service facility should be located to the place with the highest population potential of the second iteration. Locating facilities through the iterations should continue until no more facilities can be located within the ideal distance (e.g., 2.5 km for local or 10 km for subregional facilities). The next iterations continue searching for a place using the same approach but with higher service area distance, 𝑀𝑃𝑖𝑚𝑎𝑥 (for example, 5 km for local services).
A
M
6)
N
U
5)
(Eq. 1)
A
CC
8)
At the same time, the procedure should start searching for the best location for facilities with lesser service area population. The service area population that serves to these facilities cannot be less than the minimum service area population (e.g. 5.000 for local and 50.000 for subregional facilities). In countries and within remote islands below certain population, at least one facility should be allocated. For instance, Cyprus and Malta have a total population that is below the population requirements to have a regional facility, but they should still have one regional facility. The iterations should be terminated after locating all the facilities to the locations with the highest population potentials available within the determined ideal and maximum service area distances and with ideal and minimum service area populations. There might be some remote regions, which do not satisfy the thresholds for minimum service area population and maximum distance to reach them. No facilities for them within the determined distance limits. People are supposed to travel further distances to reach to the closest facility in these regions, e.g. more than 5 km for a local services. This approach will ensure capturing real life patterns in sparsely populated regions. Once all the facilities are allocated, compute road distance (𝐷𝑛𝑒𝑎𝑟𝑖 ) and (driving) travel time (𝑇𝑇𝑛𝑒𝑎𝑟𝑖 ) to the nearest facility for each cells / grids. A simplified approach could rely on the straight-line distances. Finally, calculate average distance (𝐴𝑣𝑟𝐷𝑖𝑠𝑡𝑘 ) and average travel time(𝐴𝑣𝑟𝑇𝑖𝑚𝑒𝑘 ) to the nearest facility per person for a region (k), e.g., NUTS0, NUTS3 regions. In order to make this computation, the road distances and travel times, calculated for each cells in step 9, should be weighted by the population (𝑃𝑖 ) of those cells as following:
EP
7)
(Eq. 2)
TE
D
𝑀𝑃𝑖𝑚𝑎𝑥 = ∑𝑗 𝑃𝑖𝑗 𝐷𝑖𝑗 −1 𝑤ℎ𝑒𝑟𝑒 𝐷𝑖𝑗 ≤ 𝑡ℎ𝑒 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑓𝑎𝑐𝑖𝑙𝑖𝑡𝑦
9)
10)
𝐴𝑣𝑟𝐷𝑖𝑠𝑡𝑘 = (∑𝑖 𝑃𝑖 𝐷𝑛𝑒𝑎𝑟𝑖 )/ ∑𝑖 𝑃𝑖
(Eq. 3)
40
𝐴𝑣𝑟𝑇𝑖𝑚𝑒𝑘 = (∑𝑖 𝑃𝑖 𝑇𝑇𝑛𝑒𝑎𝑟𝑖 )/ ∑𝑖 𝑃𝑖
(Eq. 4)
The outcome of these equations can be used as a measure of accessibility to generic services, i.e., average road distance to the nearest local facility (per person / km) or driving time to nearest local facility (per person / minutes).
MAE
MAPE
SEAD
Correlation coefficient - grid level comparison
Correlation coefficient - LAU2 level comparison
14.3 km
6.1 km
110 %
57%
0.40
0.53
20.0 km
22.1 km
8.3 km
94%
70%
0.45
0.54
9.7 km
11.3 km
4.2 km
58 %
71%
0.55
0.56
9.6
-30%
4.8 km
6.4 km
1.6 km
559
Estonia
7.2
9.0
25%
9.3 km
8.1 km
-1.2 km
227
France
13.4
12.2
-9%
4.6 km
4.3 km
-0.3 km
United Kingdom
10.7
12.3
15%
4.1 km
2.7 km
Belgium
15.1
13.8
-8%
3.1 km
2.8 km
M
Netherlands
24.2
12.9
-47%
2.2 km
Greece
11.5
10.9
-5%
6.2 km
Latvia
7.1
9.6
35%
10.1 km
EU8 Average
13.3
12.2
-8%
4.3 km
Health Facilities
U
13.6
A
Lithuania
N
Difference in average distance
Simulated Facilities
Health Facilities
Aggregation, first at the level of communes (LAU2) using 1km population grids, then at country level using communes4
Number of communes
Aggregation at country level using 1km population grids
SC RI PT Simulated Facilities
8.7 km
Average distance per person
Difference (%)
Simulated facilities per 100 000 inhabitants
Country Name
Health facilities per 100 000 inhabitants
Table 4: Average distance per person to the nearest facility – observed vs. simulated results
9 156
5.4 km
3.4 km
2.3 km
38 %
75%
0.68
0.61
-0.3 km
589
3.9 km
3.9 km
1.1 km
24 %
87%
0.61
0.65
2.3 km
0.1 km
418
2.7 km
3.0 km
0.6 km
21 %
95%
0.69
0.65
5.7 km
-0.5 km
1 021
16.4 km
18.0 km
6.6 km
68 %
75%
0.76
0.78
7.7 km
-2.4 km
119
21.2 km
17.6 km
4.7 km
26%
87%
0.79
0.80
3.6 km
-0.7 km
48 544
9.0 km
9.9 km
3.9 km
54 %
73%
0.67
0.65
A
CC
EP
TE
-1.4 km
D
36 456
4
observedi −simulatedi
MAE: Mean Absolute Error = ∑𝑛𝑖=1 |
MAPE: Mean Absolute Percentage Error =
𝑛 100 n
n
∑
i=1
|
observedi −simulatedi
|
observedi
|
SEAD: Share of zones with an absolute error less than half of the Average Distance in the country. More explicitly, the number of zones in a country with an absolute error smaller than the half of the country’s average distance, divided by the total number of zones and multiplied by hundred.
41