Sunteți pe pagina 1din 22

Estimation of distance-decay parameters - GIS-based

indicators of recreational accessibility

Hans Skov-Petersen
Danish Forest and Landscape Research Institute
Department of Urban and Regional Planning
Hrsholm Kongevej 11
DK-2970 Hrsholm
Denmark
Hsp@fsl.dk

Abstract. Distance-decay - as a quantitative conception of the phenomena


that things being further away are less likely to be used has been used as
one of the base assumptions when modelling human spatial behaviour,
including assessment of accessibility of resources and mobility of the users.
The article presents and discusses a number of approaches to distance-decay,
as basis for calculation of environmental indicators as well as to serve as
parameters to numeric models. Based on a concrete case of car-born
recreational behaviour in Denmark, it is demonstrated how distance-decay
parameters can be estimated and implemented in a GIS-context. Further on,
procedures for setting digital topological networks up for calculation of
indicators of accessibility and mobility are given. Finally, a number of
resulting indicator-maps are shown.

Keywords: GIS, recreation, accessibility, mobility, distance-decay, indicators

Introduction

The location of the resources relative to the users, the transport-system, and the way
space and distances influences the potential usage of facilities are central issues
when modelling accessibility and mobility. The influence of distance on potential
usage is often conceptualised through distance-decay functions (e.g. Hansen, 1959,
Koenig, 1980 or Fortheringham, 1981), expressing the way increasing distance or
travel-cost has an inverse effect on the possible usage, i.e. it is less likely that
facilities far away is used that those at closer range.
Using GIS for accessibility-modelling ahs a number of advantages, including: a) it
provides an easy assessment of transportation options, as represented by a digital
road network, b) data can be handled in a generally more flexible way, including a
wider range of options for integration of data from different sources, and finally c)

GIS enables cartographic presentation of results, which again opens for visual
interpretation and error assessment.
It is the purpose of this paper to:
Discuss the various ways models of interaction, including accessibility and
mobility, have implemented distance-decay,
to discuss the performance of different approaches to distance-decay,
to demonstrate how distance-decay parameters can be estimated from empirical
data for a concrete case of car-borne recreational activities in Denmark, and
finally
to present cartographic results of models based on the concept of distancedecay.

Accessibility, mobility and indicators

When addressing issues related to human spatial behaviour, including land use- and
transportation-planning, the assessment of accessibility and mobility often comes
into question. The definition of the two terms is non-trivial and has been addressed
by a vast number of authors since the emergence of the early attempts of the
interaction modelling (Hansen, 1959). A comprehensive attempt to define and
classify accessibility and mobility is beyond the scope of the present text. Interested
readers are advised to consult e.g. Koenig (1980), Heanue, et.al. (1995), Handy and
Niemeier (1997), or with special reference to access to the landscape, Skov-Petersen
(forthcoming).
A first very coarse, distinction can be made between spatial/physical accessibility on
one side and social accessibility and mobility on the other. Spatial or physical
mobility is the ability to physically move in space and achieve goals or objectives at
a transport-distance or time from the origin. For the purpose of the present paper a
spatial, physical approach to accessibility and mobility is taken. Social mobility
the option for changing socio-economic status (Johnston, 2000) is not considered.
Modelling spatial/physical accessibility and mobility is aiming at a quantitative
assessment of the transportation options with respect to the potential use and/or the
potential need for specified locations. This way accessibility becomes a measure of
spatial opportunities (Cervero, et al., 1999) rather than actual behaviour or wishes.
Both indicators of accessibility and mobility must include a notion of at least two
parameters: the mode of transport, including the way the transport system influences
human activity and a localised quantification of the resource made available and/or
potential need (e.g. Wickerman, 1974 and Geertman and V. Eck, 1995). This can be
further developed to include three components: the transport system, and a
distinction between the location of the facilities and the location of the users (e.g.
Handy and Niemeier, 1997 or Johnston, 2000). The concepts of the negative
influence at increasing distances on human behaviour often referred to as distancedecay is discussed overleaf in this text.

An attempt to define mobility could be the ease by which a person at a specified


location can reach and make use of a certain facility (Kronbak, 2000). This way
mobility must be seen as an attribute to the point of departure, e.g. the dwellings of
the people under consideration (Heanue, et al., 1995). The ease or cost to reach the
facility can be assessed in time, money, distance etc. Measures of mobility can be
aggregated or disaggregated. The distance or travel-time to the closest school from
a dwelling is an example of a disaggregated mobility-measure. If all the available
facilities within a specified distance or cost are considered, an aggregated measure
occurs. This way, potential resource can be used, as an indicator of aggregated
mobility, e.g. the available area of parks within 30 minutes of walk.

Fig. 1. Potential available resource an indicator of mobility: Measured in terms of


the available resource which persons at a specified location can potentially reach..
Accessibility, on the other hand can be defined as the ease by which a facility can
be reached from one or more locations or points o f departure (Kronbak, 2000).
Hence, accessibility becomes an attribute to the facility (Heanue, et.al. 1995). As
mentioned in the context of mobility, the ease or impedance can be measure in time,
money, distances etc. Again a set of aggregated and disaggregated measures can be
described. Morris, et.al. (1978) uses the term relative accessibility for disaggregated
accessibility. The same authors refer to aggregated measures as integral
accessibility. An example of disaggregated accessibility could be how well a
museum can be reached by the children from a specified school. An aggregated
accessibility indicator could be the total number of children that can reach a
specified museum within 30 minutes drive by public transport.

Fig. 2. Potential usage an indicator of accessibility. Measured in terms of the ease


by which a facility can be reached from one or more points of departure.

The distinction between accessibility and mobility - the potential usage and resource
is not always made. Very often the term accessibility is used for both
phenomenas. An example is given by Cervero et al. (1999) who addresses both
housing accessibility and job accessibility. The first, being an attribute to the
residential areas, would according to the discussion above be seen as an indicators or
measure of mobility. The latter is an attribute to the employment areas, and can
accordingly be seen as an indicator or measure of proper accessibility, as it will be
used for the remainder of this text.
The use and generation of Indicators are often motivated by the need for monitoring
the state of the environment on an even basis over time and different regions. In the
introduction chapter of the Europes Environment the Dobris Assessment
(Stanners and Bourdeau, 1991, p 6) a call is made for the generation of a set of
descriptive indicators for the production of environmental quality profiles enabling
comparison the environmental status between regions. Further the indicators should
help explaining the changes over time. To facilitate these two objectives, indicators
should catch the essence of the state of the environment in a standardised and
communicative way. An example relevant in the present context, is the use of local
access to green areas within 15 minutes walk as an indicator of urban environment
(Stanners and Bourdeau, 1991, p 264).
Accordingly the distinction between assessment of accessibility and mobility in
terms of indicators and quantitative models of phenomenas based on empirical
measurements becomes evident. Indicators are meant to have comparability and
communicativity as core characteristics whereas quantitative models attempt to
describe the phenomena as accurately as possible.

Formulation of distance-decay functions

Tobler (1970) (cited in Johnston, 2000, p. 182) formulates a first law of geography:
everything is related to everything else, but near things are more related than
distant things. The attenuation of a pattern or process with distance a phenomena
which is of course intuitively and empirically easily recognised and broadly
accepted is often referred to as distance-decay. Other terms used for the same
include Impedance Function (Koenig, 1980, Handy and Niemeier, 1993) and
Distance Lapse Rate (Johnston et al. 2000).
A generalised formulation of potential accessibility or mobility, based on the mass
(facilities or needs) surrounding sites including distance-decay is shown below
(formula 1).

Pi = M j * f (d ij )
n =1

Where:
Pi
is the potential accessibility or mobility of place i,
Mj
is the mass, e.g. population of place j,
f()
is the distance-decay function, and
dij
is the distance or cost between target i and origin j
Formula 1: Calculation of potential using a generalised distance decay function.
In the case of a sharp threshold masses of points being further away than the given
threshold are simply not included in the calculation the decay function f(d ij) can be
formulated as formula 2. This special case of distance-decay is referred to by Koenig
(1980), and Cervero et.al (1999) as isocronic definition, and cumulative
opportunities measure by Handy and Niemeier (1997).
f(dij)
f(dij)

=1
=0

for
for

dij
dij

<=
>

threshold, e.g. 30 minutes


threshold, e.g. 30 minutes

Formula 2 : Example of distance decay function as a sharp threshold.


Inspired by the Newtonian theories of gravitational attraction a classically
implementations of a distance decay are power functions (formula 3). Other writers
including Fortheringham (1981, pp 425) broadens up the term gravity model to
include any aggregated spatial interaction model in which interaction volume is
a function of nodal propulsiveness, nodal attractiveness and distance.

Pi =

Mj

j=1

ij

Where:
Pi
is the potential number of e.g. people attracted to town i,
Mj
is the mass, e.g. populations of the town j,
dij
is the distance between i and j, and

is the exponent of the potential function.


Formula 3 : Calculation of potential using a power distance-decay function.
Alternatively an exponential function can be used to represent the distance decay.

Pi = M i * exp( * d ij )
j =1

Pi
Mj
dij

Where
is the potential number of e.g. people attracted to town i,
is the mass e.g. populations of the town j,
is the distance between i and j, and
is the exponent of the function.

Formula 4 : Calculation of potential using a exponential distance-decay function.


The estimation of the exponent parameter is crucial. The higher, the more abrupt is
the cut of of influence of populations being situated at greater distances. As stated
by Fortheringham (1981, p 425) A distance-decay parameters measures the
relationship between observed interaction patterns and distance when all other
determinants of interaction are constant. The parameter is estimated as a best-fit to
the current situation, represented by an empirical data-set. Distance-decay, both in
terms of the function involved and the parameters, varies between different regions,
for different activities, and different modes of transport. In his now classical study
Hansen (1959) list findings of parameters-estimates of 2.0+ for school trips, 2.0 for
shopping trips, 1.1 for social trips and 0.9 for work trips. This indicates that the
average radius of activity for interaction between home and work is greater than
between home and school, which corresponds well with what is intuitively expected.
Several writers including Koenig (1980), Davidson, (1997) and Dalvi and Martin
(1976) regards Hansen (1959) as one of the first definitions and formulations of
accessibility indicators. The latter even refers to an accessibility index based on a
power-distance-decay function as .. the Hansen type of measure (p 18 and p 19).
According to Fortheringham (1981) estimates of distance-decay parameters are
functions of spatial structure as well as interaction behaviour. Hereby the modelling
of spatial behaviour is not just as a consequent of the available transport-system
which again is, at least partly, a function of the local topography etc. - but also
social differences etc. Hence, distance decay-parameters will be different for
different regions. Figures given for residential exchange between American cities
shows to vary from -0,01 in the area of Chicago indicating that people tend to
move very short - to its highest values of -2,3 and more in a NW/SE band across the
central Rocky Mountain area (indicating very long exchanges of dwelling). In SE
USA parameters between -1,4 and -1,9 are found. It is not surprising that people
move longer in scarcely populated areas than in areas of higher population densities.

The reason why the parameters provided by Fortheringham (1981) are all negatives is that
the interaction measure used is multiplying the mass-variable with the distance variable
(raised to the power of decay-parameters). In the Hansen (1959) study referred above a
measure like formula 3 is used, which leads to positive distance-decay parameters.

The issue is, that when reusing distance-decay parameters in other geographical
areas care should be taken.
Further more when comparing decay parameters arising from different sources care
should be taken that the units of the original empirical data are comparable. Even
though the same decay-function is used, estimated parameters based on measures in
miles, minutes and in km will obviously not be the same.
When it comes to reaching better description and explanation of imperial data higher
order approaches including power and exponential decay-function have an
obvious advantage. The precision of the model and hereby the resulting prediction
will in many cases be better if a better fit to the in-going data is obtained. On the
other hand, the results of higher order models might not be as easily communicated
as can more simple measures. Example given the number of people that can reach a
location within half an hour being a simple, sharp threshold or isocronic measure
is readily understood whereas a population-potential including a higher order
distance-decay is much harder to comprehend by laypersons (for example of this
discussion see Koenig (1980), Geertman, 1996, Handy and Niemeier, 1997, or
Skov-Petersen, 1998). Further care should be taken if no empirical data are available
for estimation of parameters for higher order decay functions. The implementation
of standard values e.g. 1.8 as the -value in power models seems hard to justify
especially when considering the above discussion of distance-decay parameters
sensitivity to differences in geography, activity and unit of cost. A sharp threshold
approach must be considered because of the advantage of enabling an easier
understanding of the background of the model and hence the results.

Fig. 3. Potential forest resource (an indicator of mobility) within 5 min drive by car.
Please note that the threshold of the legend is based on quartiles, i.e. the first class
out of ten contains the 10 % of the number of polygons having the lowest values
etc. 2Therefore and because the potential resource obviously are different for the
different time bands (5, 15, and 30 minutes) - values are not directly comparative
between figures 2, 4 and 5. The same goes for the figures 11-14.
The selection of time restriction in a sharp threshold-model is obviously important.
As discussed earlier the time restriction ought to reflect an empirical knowledge. On
the other hand, there is also a scale factor including the type of phenomena and the
spatial configuration of the region or area under consideration. On one side if the
time restriction becomes too big related to the extend of the geographical region
and the underlying data (forests and population) are more or less evenly distributed
over the area the results will be just concentric rings reflecting the shape of the
area. E.g. the case of 30 minutes drive (fig.5): Roughly speaking 30 minutes drive is
the radius of the island of Zealand. As can be seen - and what is not surprising the
mobility is biggest close to the centre of the island, i.e. most of the island can reach
it. This way the map will reflect the underlying geography, rather than the
phenomena considered. On the other side, if time restriction becomes smaller like
in the case of 5 minutes drive (fig.3) the map will look more and more like the
original forest map. If a time band of 0 minutes were investigated the result would
be exactly like the original map. Scale, phenomena, and knowledge of behaviour

Alternatively the quintile thresholds could be set in a way that the first class represents the
polygons covering first 10 % of the area etc.

must all be kept in mind when considering sharp threshold distance-decay


parameters.

Fig. 4. Potential forest resource (an indicator of mobility) within 15 min drive by
car.

Fig. 5. Potential forest resource (an indicator of mobility) within 30 min drive by
car.
To conclude the above discussion, a rule of thumb could be to use sharp thresholds
when the measures of interaction are to be used as direct presentation of indicators
of mobility or accessibility. When the me asures are to be used as parameters in a

model for explaining and prediction the utility or available resources, a higher order
decay-functions should be considered.

Fitting distance-decay parameters for car-trips to the nature.

For the present case-study empirical data obtained from household interviews was
used (Jensen, 1999). The interviewed persons were asked how much time they spend
and which means of transport they used last time they had recreation activity in the
nature. The responses were given in pre-set intervals, 0-2 minutes, 2-5 etc. Table 1
shows the accumulated number of persons responding that they used car, stratified
into time-intervals. Further, the numbers are accumulated, and recalculated into
relative numbers (compared to the total number of answers). Finally the
accumulated, relative figures were inverted to vary from 1 at the shortest distances
to 0 at the longest. The estimation of power and exponential parameters was
performed by non-linear regression. Estimation of isocronic, sharp threshold was
performed by fitting non-accumulated data to a gaussian distribution function.

0
2
5
10
15
30
60
90

2
5
10
15
30
60
90
150

8
76
269
343
231
116
33
33

8
84
353
696
927
1043
1076
1109

0,01
0,08
0,32
0,63
0,84
0,94
0,97
1,00

0,99
0,92
0,68
0,37
0,16
0,06
0,03
0,00

Table 1: Empirical background for estimation of distance decay-parameters. The


total number of relevant (car drivers) responses is 1109. The last column
corresponds to the measurements of fig. 6.
In fig. 6 the empirical data is plotted against a power, an exponential and a sharp
threshold distance-decay function. It appears that the exponential function gives the
best fit.

1,2

0,8

Measure
Potential
0,6
Exponentiel
Threshold

0,4

0,2

0
0

20

40

60

80

100

120

140

160

Fig. 6. Example of revealed behaviour in terms of time spent when going by car
from the home to the nature (Jensen, 1999). Parameters were derived by non-linear
regression techniques, using the SAS (Statistical Analysis System) procedure NLIN
(SAS, 1999): Potential (power): = -0,346625. Exponential: = -0,049281.
Threshold: Mean = 13,36. It appears that the exponential function gives the best
numeric description of the empirical data.
The shape of the curve of fig. 6 representing the empirical data gives the impression
of a slight S-shape, which could lead to an expectation that a gausian function, or a
gausian function with an offset, could provides an even better fit. The advantage of a
gausian distance-decay is that it can take into account the phenomena that changes
in distance might be less significant both when it comes to very close distances as
well as very far. E.g. the present case of recreational behaviour: Whether there is 5
or 10 minutes transport time might not make much of a difference when selecting a
site for a walk. Neither would a change from 45 to 50 minutes. But a change
between 25 to 30 could turn out to be crucial. To that matter the gausian approach
can be seen as closer to the sharp threshold approach. One of the rare references
made to implementation to a gaussian approach to distance-decay is provided by
Vickerman (1974). For the purpose of the present text no further attempt has been
made to implement the concept and techniques of a gausian based distance-decay
function. This is partly due to the fact that gaussian distance-decay is not directly
supported by of the GIS-programmes considered.

Method and data-background

For the purpose of the present context the vector-approach to spatial data analysis
was selected. The ARC/INFO GIS-software supported by applications created by
means of the scripting language AML was used. The transport options was
represented by a road network. The population data and the nature resource were
represented by polygon coverages (see fig. 7, 8, and 9). For the aggregation of
forest- and population-information to the nodes of the network a generic routine was
developed. The routine can be used for other purposes of the same kind. There are
two basic operations involved in the calculation: Attachment of information to the
nodes of the road-network be means of polygon-to-point aggregation of background
data (e.g. population and nature resources) and calculation of accessibility or
mobility, including creation of a connectivity matrix (see fig. 10).

Population (Paris h)
1 - 27
27 - 34
34 - 45
45 - 61
61 - 89
89 - 176
176 - 515
515 - 1706
1706 - 34650

Fig. 7. Population densities (persons per


km2) of Danish parishes (1:25.000) used
as data-background for the present study.

Forest

Fig. 8. Forests of Denmark used as databackground for the study. Obtained from
1:200.000 digital vector map.

!!
!
!

! !
!

!
!

!
!

!
!
!
!
!
!

!
!
!
!
!
!

!
!

!
!
! !
!

!!

!
! ! !!!
! !
!
!
!
!
!
! ! !
!
!

!
!

! !
!
! !
!

!
!

!!

!
!
!
!

!
!

! !
!
!
!
!
!
!

!
!!
!
!

! !
! !

!
!
!
!!

!
!!
!
!
! !
!
! !
!
!
!
!
!
!
!
!!
!
!
!
! !
!
!
!
!
!
! !
!
!! !
!
!
!
!
! !
!
!
!
!
! !
!
!
!
!
! !!
!
!! ! ! !
!
!
!
!
!
!
!
! !
!
!
!
!
!!
!
!
!!!
!
! !
! !
!
! !
!
!
!
!
!
!
!
! !!
!
!!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
! !
! !
!
!
!!
! !
! !
!
!
!
!
!
!
!
!
!
!
!
!
! !
!
!
! !
!!
!
!
! ! ! !
!!
!
!!
!
!
!
! ! ! !!
!
!
! !! !
! !
!
!
!!
!!
!
! !
! !! !
!
! !! ! ! !
!
! ! !
!!!
!
!
! ! !!
!
!
!
! ! ! !
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
! ! !
! !
!
!
!
!!
!
!
!
!
!
!
!
!
!
!

!
!

!
!
!

!
!

!
!
! !
!

!
!

Facilities (forests) and needs


(population) are attached to the nodes
displayed as squares. The travel cost
in terms of driving time, calculated
from distances and speed limits are
attributed to the links (arcs) connecting
the nodes. The resulting accessibility is
assigned to the nodes. For presentation
(see the results below) the accessibility
information from the nodes is linked to
the Thiessen polygons representing
each node.

!!

! Network nodes
Ne twork li nks
Highway
Other primary road
Other road
Other road
Other road

Fig. 9. Local example (the Lilleblt bridges between the island of Funen and Jutland) of the roadnetwork used for the study (1:200.000).

The polygon-to-point aggregation is performed by creation of Thiessen polygons


around the nodes of the network and subsequently overlaying them with the
background polygon data population or forest. With respect to the forests the area
intersecting the Thiessen polygons are summed. For the population the product of
the intersecting area and the population density is added one set for each Thiessen
polygon.
The mode of aggregation is this way depending on the type of the in-going data. If
data are area independent e.g. as total population per parish they have to be
recalculated into figures relative to the area e.g. population densities (see formula
5). In the case of e.g. aggregation of the available forest area for each node the
calculation is simpler a sheer addition of the areas of the forest falling with in the
shed of each node3.

This type of aggregation is referred to as interpolation of extensive variables by Flowerdew


and Green (1995) where a further discussion of the issue is found.

Xi = Aij * x j
j= n

Formula 5: Aggregation of area data,


where:
Xi
is the aggregated value of node i
Aij
is the area over the intersection of the thiessen polygon of node i and the in
going polygon j and
xj
is the area independent value (e.g. population density) of the in going
polygon j. xj will always be 1 in the simple case of collection of areas (e.g. in
the present case of forest resources).
In principle, a connectivity matrix represents connections - as distances or costs along the cheapest or shortest route of all combinations of nodes in a network. In a
bi-directional network4 where all nodes (n) can reach each other the number of
elements of the connectivity matrix will be n2 n. In a unidirectional network the
number of combinations will be (n2 n)/2. To reduce the resulting data-set many
systems facilitates the possibility of o nly making calculations for nodes within a
specified maximum cost or distances (e.g. 1000 metre). If implemented, a distance
decay function can subsidiary be applied to connectivity matrix.

In a bi-directional network the impedance along one link (the line between two nodes, say a
and b) differs whether going from a to b or vice versa (Hagget and Chorley, 1969)..

Connetivity matrix
4
Urban areas

Road network

2
Thiessen polygons

Forests

Fig. 10. Method for aggregation of polygon data to network nodes. For details, please refer to
the text below.

1. Creation of Thiessen polygons around the nodes of the road network.


2. Aggregation of the background data (population data and nature areas), based on
the overlay with the Thiessen polygons. The calculated data are linked back to
the nodes of the road network.
3. Calculation of a distance or impedance matrix e.g. the distance between all
nodes that are less than 1000 meters apart.
4. Summarising the data from nodes connected by the impedance matrix. And
finally linking back the summarised data to the nodes of the road network.
In most cases at least in the context of ARC/INFO step 3 and 4 are an integral
part of the command used (ACCESSIBILITY) and therefore not directly issued by
the operator. Both sharp thresholds, potential and exponential distance-decay
functions are supported this way. When processing in ARC/INFO, sharp
threshold cases, the -parameters is simply set to 0, meaning that space has no
influence. The threshold is controlled by a maximum-impedance parameter set
elsewhere. In a more rare case, like when implementing a gausian distance decay
function the connectivity matrix has to be build explicitly, e.g. by using the
NODEDISTANCE command.

Results

Fig. 11 and 12 show cartographic representations of results from modelling


accessibility in Denmark, based on the distance-decay parameters estimated in the
previous section. As was earlier stated, accessibility is an attribute to facilities, e.g.
natural areas. The maps must interpreted as the number of people that can reach a
location, given the provided distance-decay. It appears that it is much easier to
communicate results of sharp threshold calculation at this spot 24351 persons can
come within 15 minutes drive from their home than is results from higher order
decay functions, including exponential. Even though both maps are based on the
same empirical data and even though numbers are situated approximately in the
same range, they appear quite different. Fig. 11 the sharp threshold case displays
a much more contrast-rich picture than the exponential case. This corresponds quite
well with what is expected from looking at the graphs of fig. 7: Locations that are
generally slightly more than 15 minutes away from areas of high publication wi ll
tend to be underestimated compare to the exponential case. On the other hand
accessibility will be overestimated if the general distance to populated areas are
slightly less that 15 minutes. It must be concluded that even though the exponential
case gives a better fit to empirical data the message comes true more easy for the
sharp threshold map.

Number of people
0 - 3982
3982 - 5785
5785 - 7258
7258 - 8787
8787 - 11044
11044 - 13904
13904 - 17579
17579 - 24611
24611 - 42985
42985 - 230727

Fig. 11. Denmark: accessibility (sharp threshold = 15 minutes drive by car). To


enhance contrast the classes of the legend are quintile, i.e. the number of units of
each legend class is even. Figures are comparable to fig. 12, even though thresholds
between classes are not the same.

Number of people
0 - 18428
18428 - 26949
26949 - 34204
34204 - 41739
41739 - 48843
48843 - 56945
56945 - 65074
65074 - 76725
76725 - 99696
99696 - 214021

fig. 12. Denmark: accessibility (exponential, = -0,049281). Note that the classes of
legend are quintile

Sqr. Km. of forest


< 11
11 - 19
19 - 26
26 - 31
31 - 37
37 - 43
43 - 51
51 - 60
60 - 75
> 75

Fig. 13. Denmark: Potential resources (km2 forest) as indicator of mobility (sharp
threshold = 15 minutes drive by car). To enhance contrast the classes of the legend
are quintile, i.e. the number of units of each legend class is even. Figures are
comparable to fig. 14, even though thresholds between classes are not the same.

Sqr. Km. forest


< 74
74- 95
95 - 113
113 - 130
130 - 145
145 - 161
161 - 178
178 - 202
202 - 246
> 246

Fig. 14. Denmark: Potential resource (km2 forest) as indicators of mobility


(exponential, = -0,049281). Note that the classes of legend are quintile.
Fig. 13 and 14 shows cartographic presentation of indicators of mobility calculated
on the basis of estimated distance-decay parameters from the previous section. The

most apparent different between the two maps is that fig. 13 based on a sharp
threshold indicator appears much less smooth and more contrasted than the one
based on the exponential decay-function (fig. 14). The maps show the potential
resource in terms of km2 forest with respect to one or the other distance-decay
function and should be read something like If I am living at this location . I
would have this amount of forest at my facility. Again as above it appears that a
sharp threshold indicators is more readily communicated and understood than the
one based on exponential distance-decay function.

Conclusion

A set of different approaches to distance-decay has been discussed and


demonstrated. The quality or success of the model-results can be judged from two
different angles: a) its communicative qualities as in the case of environmental
indicators and b) as a quantitative description of phenomena, based on empirical
data. It has been argued and demonstrated that a sharp threshold approach to
distance-decay reveals more easily understood results making them more useful in
the context of e.g. decision-making involving laypersons , politicians or other not
having sufficient technical or professional background to grasp results from models
based on higher-order distance-decay functions. If the aim is as numerically accurate
as possible to describe a phenomenon, based on higher-order decay-functions can be
implemented. But it requires that relevant empirical data is available. Human
activity and hereby the functions and parameters describing them varies over time,
space and between different activities. Use of higher-order decay-functions for
generation of indicators or map of larger areas or representing non-specified
activities therefore makes no sense. If the objective is to generate knowledge of what
will happen over time if conditions changes, the predictive accuracy of the model is
required, calling for more complex modelling approaches, including higher-order
distance-decay functions.
There is a need for both indicators of accessibility and mobility. The potential
number of users, as an indicator of accessibility, is needed by the managers of the
facilities. E.g. it is necessary for the forest manager to obtain estimates of the
potential number of visitors. The urban planner on the other hand, needs estimates of
the green resource for different parts of the urban area under consideration.

References
Cervero, R., Rood, T. and Appleyard, B. 1999. Tracking accessibility: employment
and housing opportunities in the San Francisco Bay Area. Environment and
Planning A, Vol. 31. Pp 1259-1278.
Dalvi, M.Q. and Martin, K.M. 1976. The measurement of accessibility: Some
preliminary results. Transportation 5. Pp. 17-42.

Davidson, K.B. 1977. Accessibility in transport/land-use modelling and assessment.


Environment and planning A, Vol 9. Pp 1401-1416.
Fortheringham, A.S. 1981. Spatial structure and distance-decay parameters. Annals
of the Association of American Geographers. Vol. 71, No. 3. Pp. 425-436.
Geertman, S. and Van Eck, J.R.R. 1995. GIS and models of accessibility potential:
An Application in Planning. International Journal of Geographical Information
Systems. Vol. 9. No. 1. Pp. 6780.
Hagget, P., and Chorley, R. 1969. Network analysis in Geography. Edward Arnold.
Handy, S.L. and Niemeier, D.A. 1997. Measuring accessibility: an exploration of
issues and alternatives. Environment and planning A. Vol. 29. Pp. 1175-1194.
Hansen, W.G. 1959. How accessibility shapes land use. Journal of the American
Institute of Planners. Vol. 25. Pp. 73-76. Reprinted with permission by Kraus
reprint corporation (1966).
Heanue, K., Menchkhoff, G., Peyrebrune, H. and Pisarski, A. 1995. Urban
mobility: An international perspective. Routes Roads Special 1. PIARC World
Association.
Jensen, F.S. 1999. Pers.Com. Frank Sndergrd Jensen, Danish Forest and Landscape
Research Institute. Fsj@fsl.dk. General description of the project supplying the
data is provided by Jensen, F.S. 1999. Forest recreation in Denmark from 1970s to
the 1990s. The Research Series, Nr. 26. Danish Forest and Landscape Research
Institute. Hrsholm kongevej 11, Dk-2970 Hrsholm. Denmark.
Johnston, R. J., Gregory, D., Pratt, G. and Watts, M. (eds.) 2000. The dictionary
of human geography, 4th edition. Blackwell Publishers Ltd.
Koenig, J.G. 1980. Indicators of urban accessibility: Theory and application.
Transportation 9. Pp 145-172.
Morris, J.M., Dumble, P.L. and Wigan, M.R., 1978. Accessibility indicators for
transport planning. Transport Research A. Vol. 13A. Pp. 91109.
SAS, 1999. SAS users manual. Statistical Analysis Systems.
Skov-Petersen, H. 1998. GIS-analysis of barrier effect of larger traffic constructions.
GIS planet 1998. Lisbon.
Skov-Petersen, H. 2001. Indicators of accessibility and Mobility, with special
reference to recreational behaviour in the landscape. Forthcoming.
Stanners, D. and Bourdeau, P. (eds). 1995. Europes Environment The DOBRIS
Assessment. European Environment Agency.
Vickerman, R.W. 1974. Accessibility, attraction, and potential: a review of some
concepts and their use in determining mobility.

S-ar putea să vă placă și