Documente Academic
Documente Profesional
Documente Cultură
density grids
F.J. Gallego,
Institute for Environment and Sustainability, JRC, Ispra (Italy)
Summary
We discuss several possible criteria to classify the European Union (EU) into
rural and urban areas. Some unexpected results are highlighted with definitions
based only on population density per administrative unit or only on land cover
type. These abnormal results come partly because the definitions depend too
strongly on the size of the communal territory.
In the approach we propose as a basis for discussion, the classification units are
the communes, but a major role is played by urban agglomerations, that are
defined and determined in a way that does not depend on administrative
boundaries. The classification proposed has three major groups: urban, semiurban and rural. Each of them has been subdivided, but different criteria for subclassification have been prepared.
OECD definition.
The OECD has given a definition of rural areas based on the percentage of the
population of a region living in rural communes (OECD, 1994). A commune is
classified as rural if the population density is below 150 inhabitants per km2. This
definition has the merit of being easily applicable, but has some limitations:
The commune classification into urban/rural depends too much on the
area of the commune. Let us put an example to illustrate some awkward
effects of this definition: In Extremadura (Spain), this definition gives only
6 urban communes, out of which 4 have less than 6,000 inhabitants. The
main cities of the region (Badajoz, Cceres and Mrida) are classified as
rural because their communes include wide areas of woodland and shrub.
It does not take into account the characteristics of the surrounding area, in
particular if it belongs to the outskirts of a big city. Figure 1 shows the case
of Aldea de Trujillo, classified as urban because the communal territory
only includes the small urban nucleus (439 inhabitants in 0.3 km2), without
taking into account that it is surrounded by natural and agricultural areas.
Some relatively large towns are labelled as rural because the communal
territory contains large empty areas (table 1). More than 250 communes
above 20,000 inhabitants have a density < 150 and would be considered
as rural with the OECD definition. Most of them have a fairly large urban
nucleus. If the commune is used as classification unit, a specific category
might be meaningful for this type of communes containing an urban
nucleus and large agricultural or natural areas.
Population
Aldea de Trujillo
Calamonte
Puebla de la calzada
Valle de Santa Ana
439
5564
5480
1338
Area
commune
(ha)
35
770
1419
376
Density
1272
723
386
356
Population
Badajoz
Cceres
Mrida
122225
74589
49284
Area
commune
153434
175139
86785
Density
80
43
57
Jerez de la Frontera
Uppsala
Albacete
Linkoeping
Badajoz
Oerebro
Norrkoeping
Vaesteraas
Joenkoeping
Boraas
Area
commune
(ha)
141180
246459
124310
143086
153080
184006
149073
95643
148512
118039
Population
commune
183316
167508
130023
122268
122225
120944
120522
119761
111486
101766
density
130
68
105
85
80
66
81
125
75
86
mainly urban regions: less than 15% of the region's population lives in
rural communes.
Figure 2: Rural and urban NUTS3 units according to OECD definition. (Scotland,
East Germany and islands missing).
The criterion is reasonable but has some limitations and produce unexpected
results that depend on the delimitation of communal boundaries, in particular for
large, heterogeneous NUTS3 units, for which attributing the same label of rurality
to the whole NUTS3 unit may be unfair. For example the only province in Sicily
that turns out to be mainly urban is Ragusa, while Stockholm is relatively rural.
363
109
1404
715
2973
705
4839
743
96
1672
15
209
551
762
1135
16291
2310
443
4421
1152
11727
10216
15285
3759
2090
6727
67
398
2404
12902
3910
77811
16
25
32
62
25
7
32
20
5
25
22
53
23
6
29
21
217
520
2906
183
886
157
1614
124
70
2462
13
737
313
114
1155
11471
13
26
26
39
27
7
27
24
7
33
54
54
22
32
23
29
AT
BE
DE (West)
DK
ES
FI
FR
GR
IE
IT
LU
NL
PT
SE
UK (No Scotland)
total
762
242
2471
794
5921
725
8813
1664
1597
3115
46
229
864
1004
2981
31228
64
258
1329
83
405
12
686
64
41
1424
10
420
105
41
547
5489
2310
443
4421
1152
11727
10216
15285
3759
2090
6727
67
398
2404
12902
3910
77811
33
55
56
69
50
7
58
44
76
46
69
58
36
8
76
40
217
520
2906
183
886
157
1614
124
70
2462
13
737
313
114
1155
11471
29
50
46
45
46
8
43
52
59
58
77
57
34
36
47
48
AT
BE
DE
DK
ES
FI
FR
GR
IE
IT
LU
NL
PT
SE
UK
Total
Mainly
urban
19
55
38
12
25
n.a
46
36
n.a
46
n.a
56
25
n.a
57
45
NUTS3
Relatively
rural
28
55
55
57
49
19
57
61
75
51
70
57
24
33
75
54
Mainly
rural
34
40
55
71
55
6
56
43
76
49
n.a
69
40
7
89
33
average
33
52
52
65
50
7
56
44
76
49
70
56
36
8
65
42
Table 6: % of the territory used for agriculture (UAA) in rural/urban NUTS3 with
the OECD definition
AT
BE
DE
DK
ES
FI
FR
GR
IE
IT
LU
NL
PT
SE
UK
Total
Mainly
urban
12
31
22
12
10
n.a
28
4
n.a
25
n.a
53
19
n.a
24
25
NUTS3
Relatively
rural
14
30
31
49
26
19
33
23
10
28
28
53
15
27
31
29
Mainly
rural
16
9
31
64
27
6
29
20
4
27
n.a
61
26
5
1
17
average
16
25
33
59
25
7
31
20
5
27
28
53
23
6
22
22
Argenton-Chteau (Deux-Svres)
~1000 inhab, 49% artificial (CLC)
Country
IT
ES
ES
ES
IT
FI
PL
ES
ES
ES
ES
SI
Population*
(1000 inh.)
3089
847
693
688
675
551
432
413
404
388
380
379
Area
(km2)
1505
135
1065
396
239
685
301
209
1257
198
888
275
% artificial in
CLC
27
29
6
15
25
15
30
22
6
14
4
24
Table 8: some cities with less than 30% artificial land in CLC.
* Census data were not available for this work. Population data have been derived from the
Landscan raster layer. These figures are not very inaccurate and should be seen as an indication
of the order of magnitude.
This classification can be nuanced by separating several groups of artificial land
cover types. CLC nomenclature contains 11 artificial classes, and some of them,
such as airports, mines, dump sites and leisure facilities, are not necessarily
characteristic of an urban environment.
Figure 4 shows some examples of communes that do not belong to any urban
agglomeration), but have a significant area for an airport, leisure zone, dump site
or mines.
Vintirov (CZ)
36% dump site,
15% mine
12% industrial
The 1 ha grid obtained from 1991 commune data is more accurate, but the
coverage incompleteness is a serious limitation. With this layer we have applied
a series of filters to define urban agglomerations:
Smoothing by averaging values in a circle of radius 500 m. Water was
excluded from this operation by masking with CLC.
Applying a threshold of 500 inh/km2.
Majority filter with a 11x11 moving window to smooth the shape and
eliminate small centres (with mask).
Buffering with a 5x5 moving window (no mask).
Converting to polygon shape.
Computing the total population of each polygon. Selecting polygons >
5000 inhab.
The operation was slightly different with the 2002 Landscan layer in lat-long coordinates to adapt for the coarser resolution. Steps were:
Converting the population layer (estimated number of inhabitants per cell)
into a density layer dividing by the area of each cell (varying with latitude).
threshold of 500 inh/km2
majority filter 3x3 cells (with coast line as mask)
Buffer by maximum in a 3x3 moving window (no mask)
Converting to polygons
Estimating the population of each urban polygon by summing the values
of the grid cells inside the polygon.
Selecting nuclei > 5000 inh. (notice that the population estimated from the
grid layer is not very accurate, in particular for small communes).
Urban agglomerations are classified by classes of size. A total of 3886
agglomerations have been identified counting for a total of 287 million inhabitants
For the small size classes, the number of nuclei identified depends very strongly
on the characteristics of the population grid. A first visual inspection suggests
that the population of rural centres below 20,000 inhabitants may be strongly
underestimated. Further check is needed. Other abnormal results can be found.
Figure 5 illustrates the case of Tampere (Finland) in which the algorithm
10
N. nuclei
1303
808
911
453
207
128
36
24
13
3
Total pop
9302
11109
29894
31789
28148
39301
26029
35848
45334
30142
Table 9 : number of nuclei identified > 5000 inhab. per size class.
Figure 6 shows the different results obtained from both population layers in the
Ruhr-Rhine area between Dortmund and Bonn. In Both cases the area is
identified as the largest urban agglomeration in Europe, but working with the
coarse resolution layer leads to a larger single polygon where the finer resolution
had separated several agglomerations. This is due to several reasons: different
reference date, different accuracies, different parameters chosen for smoothing
and buffering, etc., but illustrates as well that a certain subjectivity remains
always when urban agglomerations are defined, even if it is through a GIS
algorithm. Some agglomerations have been estimated above 50.000 inhabitants
from Landscan and below for the 1991 grid.
11
Figure 6: different urban agglomerations (>50.000 inh.) identified from the 1991
population grid (1 ha resolution) and the 2002 Landscan grid.
12
Rural
Peri-urban rural areas: does not intersect with any urban nucleus>5,000
inhab., but is in the area of influence of an urban agglomeration. The area
of influence has been defined through a gravity indicator (Wang and
Gouldmann, 1996), that takes into account the population of the
agglomeration.
Remote rural: distant from urban agglomerations.
5-10
10-20
20-50
50-100
100-200
200-500
500-1000
1000-2000
2000-5000
>5000
Total
Pure
urban
90
140
362
661
537
908
439
958
858
1424
6377
Mainly
Urban
80-99%
urban
agglom.
109
136
185
227
172
291
187
212
252
173
1944
Mainly
Urban
50-80%
urban
agglom.
193
209
407
369
256
352
175
255
246
160
2622
Urban with
rural area
Suburban
(periferical
urban)
991
569
576
229
83
29
5
1586
1256
1837
1340
831
860
388
375
310
212
8995
1
2483
13
14
Serrata
Candinoni
Laureana
di Borrello
Ferloleto
Galatro
6 Kilometers
15
Urban
Mixed
Rural
Purely urban
Mainly urban
Suburban
Urban with rural areas
Close peri-urban
Medium peri-urban
Far peri-urban
Medium remote
Remote
Severely remote
Unclassified
Number
communes
6377
4566
8995
2483
4871
8721
15458
27461
17839
11157
767
Total Pop.
(Landscan)
in 1000 inh.
70012
123243
52699
88782
8176
13954
22214
36230
21041
13952
765
Total
area
Km2
19960
78235
340038
547725
78248
166910
348840
800096
660613
1067734
16
Discussion
Giving an absolutely objective criterion to classify a geographical area into urban
and rural areas is probably impossible. Any method requires a choice of
thresholds, that is subjective to a certain extent. A good method should be
17
flexible, so that a potential user can easily tune thresholds for a better adaptation
to specific needs.
We have considered the commune as the unit for classification
The method we have used to classify the approximately 108,000 communes is
strongly based on the previous identification of urban agglomerations that do not
take into account communal boundaries.
Many issues have not been addressed, for example how to deal with major
tourist areas: How to consider a coastal commune that may have 2,000
inhabitants in the winter and 30,000 in August?.
We use the terms urban nucleus and urban agglomeration in a rather
unspecific and exchangeable way. It might be meaningful to classify into
relatively small nuclei, often with an approximately round shape and complex
agglomerations. A shape indicator, such as perimeter/sqrt(area), might be useful
for this purpose.
The results presented here are conceived as a basis for discussion. A number of
arbitrary choices have been made, that need to be discussed, and some
inaccuracies certainly appear, partly because of the population data grid used
and partly because of the processing.
If the approach is considered to be a valid basis, significant improvements are
needed before submitting the results to validation by Member States. Some
improvements can be made at short term, and some need a long-term work.
At short term:
Some communes have not been classified. Some of them because the
population density grid used did not cover the whole EU25: Cyprus, Malta,
Atlantic islands, and Caribbean territories were not included. Others have
not been classified because of recoverable problems in the GIS
processing.
Better tuned criterion to measure remoteness. Figure 7 suggests that the
importance of major agglomerations is too strong. Formula (1) might need
an adaptation.
The class we have called suburban communes needs further reflection.
The variety of situations included in it is too high. Many communes are
classified in this group just because they are not too far from a medium
size town.
Changing the size threshold for urban agglomerations; for example
Vanhove (1999) suggests 30,000 inhabitants, and the Buckwell report
(European Commission, 1997) considers a threshold of 50,000
inhabitants.
At long term:
18
REFERENCES
Bettio M., Delinc J., Bruyas P., Croi W., Eiden G., 2002, Area frame surveys: aim,
principals and operational surveys. Building Agri-environmental indicators, focussing
on the European Area frame Survey LUCAS. EC report EUR 20521, pp. 12-27.
http://agrienv.jrc.it/publications/ECpubs/agri-ind/
CEC-EEA, 1993, CORINE Land Cover; technical guide, Report EUR 12585EN. Office
for Publications of the European Communities. Luxembourg,. 144 pp. www.ec-gis.org
DG Agri, 2004, GIS analysis of rural areas EU25 (communal level), intermediate
report.
Dobson J.E., Bright E.A., Coleman P., Durfee R., Worley B, 2000, Landscan: a global
population databasefor estimating population at risk. Photogrammetric Engineering
and Remote Sensing., Vol. 66, N. 7, pp.849-857.
EEA, 2001, Towards agri-environmental indicators. Integrating statistical and
administrative data with land cover information. EEA topic report 6/2001,
http://reports.eea.eu.int/topic_report_2001_06/en
Gallego J., Peedell S., 2001, Using CORINE Land Cover to map population density.
Towards Agri-environmental indicators, Topic report 6/2001 European Environment
Agency, Copenhagen, pp. 92-103. http://reports.eea.eu.int/topic_report_2001_06/en
Geddes A., Flowerdew R., 2000, Geographical considerations in designing policyrelevant regions. 3rd AGILE conference on Geographic Information Science. Helsinki,
May 25-27, pp. 80-81.
Librecht I., Tirry D., Van Orshoven J., Van de Walle S., 2004, Measuring the rural
heterogeneity of NUTS 3 regions, SADL-KUL report, JRC Ispra.
OECD, 1994, Creating rural indicators. Paris.
Wang F., Guldmann J.M., 1996, Simulating urban population density with a gravitybased model. Socio-Economic Planning Sciences, Volume 30, Issue 4, December
1996, Pages 245-256
Vanhove, N., 1999. Regional Policy: a European Approach (3rd Edition ed.), Ashgate,
Aldershot.
European Commission (1997). Towards a Common Agricultural and Rural Policy for
Europe: Report of an Expert Group (Buckwell Report).
19