UVIC SSL W ORKING PAPER 05-014
MARCH 2005
Road Classification Schemes – Good Indicators of Traffic
Volume?
Eleanor M. Setton, Perry W. Hystad, C. Peter Keller
Spatial Sciences Laboratories, Department of Geography, University of Victoria
Address correspondence to Eleanor Setton, Geography Department, University of Victoria, PO BOX 3050
STN CSC, Victoria, BC, Canada, V8W 3P5. E-mail: [email protected].
We compare three road classification systems to actual traffic counts in order to assess how well
the classification systems perform as indicators of traffic volume, assuming that clear
differentiation of traffic volumes among classes is desirable. Actual traffic counts were obtained
for 215 locations in the Greater Vancouver Regional District(GVRD); the British Columbia
provincial Digital Road Atlas (DRA) and DMTI CanMap© road network provided road
classification systems. Modelled traffic volumes for the GVRD, provided by TransLink, are also
used to evaluate the classification systems. Based on the sample of actual traffic counts, we
conclude that DRA road classes provide the best differentiation of traffic volume, although within
class variation is substantial. Modelled traffic counts are not well differentiated by either the
DRA road class or subclass, indicating either poor model performance or sample bias. A
comparison of actual traffic count means for three regions in the study area with different total
population and population densities showed no spatial pattern that would explain within class
variation. Future research on within class variation is required, using a larger sample of actual
traffic counts. Overall, the use of road classes to indicate level of exposure to traffic-related air
pollution should be approached with caution, as significant exposure misclassification could
occur.
Acknowledgements: This research has been funded by the BC Centre for Disease Control, via a
grant provided by Health Canada as part of the ongoing Border Air Quality Strategy agreement
between Canada and the United States. Special thanks to all participating municipalities for
providing actual traffic count data, and to TransLink for providing modelled traffic volumes.
Key Words: actual traffic counts, modelled traffic volume, variability, exposure, GIS
Introduction
Methods for indirectly assessing personal exposure to traffic-related air pollution generally rely
on measuring proximity to roads and may incorporate some estimate of traffic volume. Where
traffic volumes are available, these numerical values can be used directly. In cases where traffic
volumes are not available, road class may be used as a surrogate for traffic volume, assuming
some hierarchy of traffic volume associated with, for example, highways, major roads, minor
roads, and local roads.
When road class is used as a surrogate for traffic volume in an exposure assessment, each road
class must represent a distinct traffic volume level in order to avoid exposure misclassification.
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
1
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
The objective of this report is to explore how well existing road classifications in British
Columbia differentiate traffic volume, and to assess opportunities for developing a method for
estimating traffic volumes in areas where no actual traffic counts exist, based on existing road
classifications. We evaluate three road classifications that are available for the province of British
Columbia (BC) – a six-level system used in the commercially available DMTI Canmap© street
network; and a five- and eight-level system, both available in the BC government’s Digital Road
Atlas (DRA) (see Table1). Actual and modelled traffic counts in the Greater Vancouver Regional
District (GVRD) are used to assess how well each classification system differentiates traffic
volume and to explore regional differences that might support the development of a model of
traffic volume based on road class.
Methods
Existing traffic counts (average daily traffic volume) at 215 locations in the GVRD were acquired
from local jurisdictions, each of which was contacted and asked to provide traffic counts for at
least 30 road segments covering a broad range of road classes. This method was deemed more
feasible than developing a random sample based on road class, given the time frame and
resources available for data collection, and the lack of data for many road segments throughout
the study area. A geographic information system (GIS) was used to create point locations for
each traffic count, and to associate the traffic count with the appropriate road segment in the
DMTI and DRA street networks. TransLink (the transit authority in the GVRD) provided GISready modelled traffic counts. Both actual and modelled traffic counts for each road class in the
GVRD, and for each road class in three sub-regions of the GVRD were compared using simple
descriptive statistics and box plots.
Table 1. Existing Road Classifications
DMTI - Class
Expressway
Principal Highway
Secondary Highway
Major Road
Local Road
Description
Usually four lanes + and very limited access to adjacent land uses
Conduits for intra-city traffic, multi-lane, large traffic capacity
Thoroughfare, large traffic capacity, generally multilane
Routes for shorter trips within the city/
Residential access
DRA – Class
Freeway
Highway
Arterial
Collector
Local
Description
Controlled access, typically divided
Primary or secondary provincial highway, single or multilane
Thoroughfare, large traffic capacity, generally multilane
Connects areas to cross town, generally one lane each way.
Residential roads
DRA – Subclass
Freeway
Highway Major
Highway Minor
Arterial Major
Arterial Minor
Collector Major
Collector Minor
Local
Description
Controlled access, typically divided
A primary provincial highway
A secondary provincial highway
Thoroughfare, large traffic capacity, more than 2 lanes
Thoroughfare, medium traffic capacity, 2 lanes (one each way)
Connects areas to cross town, more than 2 lanes
Connects areas to cross town, one lane each way.
Residential roads
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
2
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
Results and Discussion
Comparison of DMTI and DRA road classes based on actual traffic counts. A comparison of the
mean actual traffic count for each class used in the DMTI and DRA datasets suggests that DRA
‘class’ may provide better distinction of traffic volume than DRA ‘subclass’ or the DMTI ‘class’
in terms of assigning traffic-related exposure. Table 2 shows that three of the five DMTI classes
have similar means - major, highway secondary, and highway principal. DRA subclasses show
similar means for minor and major collector, and for minor and major arterial and minor
highway. The means of actual traffic counts for DRA classes suggest that DRA classes appear to
capture distinct groups of traffic volumes, suitable for indicating different levels of exposure to
traffic-related pollution.
Table 2. Mean of actual traffic counts for road classes
DMTI Class
Mean
DRA Class
Mean
DRA Subclass
Mean
Local
6,511
Local
3,976
Local
4,126
Major
15,207
Collector
8,953
Collector minor
8,580
Collector major
9,694
Arterial minor
15,321
Arterial major
17,407
Highway minor
22,242
Highway major
36,684
Highway Secondary
Highway Principal
Expressway
18,254
21,025
113,789
Arterial
Highway
Freeway
18,457
27,961
113,789
Freeway
113,789
Box plots of the traffic counts associated with the DMTI classification (Figure 1) show good
differentiation between local and other road classes, but there appears to be much less
differentiation among major, secondary highway, and principle highway classes. Notably, only
one actual traffic count was associated with a secondary highway, which may indicate that this
class is rarely used or that the provided sample is biased. Within class variability as indicated by
standard deviations is high for local and major roads (Table 3). For local roads, the mean is 6,511
and the standard deviation is 5,895. For major roads, the mean is 15,207 while the standard
deviation is 10,831. Within class variability does decline moving up the hierarchy from local
roads up to expressways. These large standard deviations support the conclusion that there is
relatively poor differentiation of traffic volumes between classes in most cases.
In contrast, the box plots of traffic counts associated with DRA ‘class’ show there are reasonable
sample sizes for local, collector and arterial classes (53, 86, and 65 respectively), but relatively
few actual traffic counts available for highway and freeway classes (Figure 2). This is not
unreasonable, as there are far fewer highways and freeways in comparison to local, collector, and
arterial roads in the GVRD. There appears to be better differentiation among classes than that
shown in the DMTI data, as there is less overlap between the boxes (representing the middle 50
percent of values) and whiskers (representing the lower 25 and upper 75 percent of values).
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
3
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
However, as shown by the standard deviations (Table 4), there is substantial within class
variation.
Boxplots of the traffic counts associated with DRA subclass (Figure 3) show little distinction
between minor and major collector and arterial roads. There is more distinction between minor
and major highways, but small sample sizes (two in each category) preclude making a definitive
statement regarding differences between these two categories. Again, standard deviations within
each subclass are relatively high (Table 5), although less so for larger roads (highways and
freeways).
140000
120000
100000
Actual Traffic Count
80000
60000
40000
146
61
157
150
69
20000
87
84
0
N=
114
Local
74
1
18
Highway (sec)
Major
7
Expressway
Highway (pri)
Figure 1. Actual traffic counts associated with DMTI road class
Table 3. Mean and standard deviation for actual traffic counts
associated with DMTI road class
DMTI Class
Mean
Standard Deviation
Local
6,511
5,895
Major
15,207
10,831
Highway Secondary
18,254
One sample
Highway Principal
21,025
9,129
113,789
22,685
Expressway
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
4
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
160000
140000
120000
100000
Actual Traffic Count
80000
60000
173
181
40000
99
98
91
20000
0
N=
53
86
Local
65
3
Arterial
Collector
7
Freeway
Highway
Figure 2. Actual traffic counts associated with DRA road class
Table 4. Mean and standard deviation for actual traffic counts
associated with DRA road class
DRA Class
Mean
Standard Deviation
Local
3,976
2,779
Collector
8,953
6,812
Arterial
18,457
10,717
Highway
27,961
10,679
Freeway
113,789
22,685
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
5
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
160000
140000
120000
Actual Traffic Count
100000
80000
60000
182
60
58
40000
20000
189
117
125
0
N=
48
40
47
15
53
2
2
7
a y j)
w
a
ee (m
Fr
y
a
)
hw min
ig
(
H
ay
hw aj)
ig
m
H
l(
)
ria
in
te
Ar l (m
)
aj
ria
m
te
Ar or (
)
ct
in
le
ol r (m
C
o
ct
le
ol
C
l
ca
Lo
Figure 3. Actual traffic counts associated with DRA subclass
Table 5. Mean and standard deviation for actual traffic counts
associated with DRA road subclass
DRA Subclass
Mean
Local
4,126
Collector minor
8,580
Collector major
9,694
Arterial minor
15,321
Arterial major
17,407
Highway minor
22,242
Highway major
36,684
Freeway
Standard Deviation
113,789
2,879
8,522
6,015
10,850
10,900
5,640
3,841
22,684
Comparison of DRA road classes based on modelled traffic volumes. TransLink provided traffic
volumes predicted by EMME/2, a widely-used traffic demand modelling software program. In
total, traffic volumes for 10,313 road segments in the Greater Vancouver Regional District were
available for analysis. The following analyses focus on DRA road class and subclass, given that
they appear to better differentiate traffic volume among classes using actual traffic counts than
the DMTI classes.
Modelled traffic volumes for each DRA class are more poorly differentiated than actual traffic
counts. Local and collector roads have similar means (4,378 and 5,377 respectively), while
freeways show lower counts than highways (Table 6). Although DRA subclass shows a more
reasonable separation of highways and freeways (means of 17,666 and 39,431 respectively), there
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
6
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
are apparent overlaps and inconsistencies for many other subclasses. For example, local and
collector minor roads have similar means, collector major and arterial minor roads have similar
means, and the mean of arterial major roads is substantially higher than that of highway minor
roads. Boxplots of the modelled traffic volumes for both DRA class and subclass (Figures 4 and
5) reflect these similarities and overlaps, and show substantial numbers of extreme values and
outliers, particularly for local, collector, and arterial roads. Also of interest are the much lower
means for freeways based on modelled volumes in comparison to actual counts – 13,546 or
39,431 (class and subclass) versus 113,789. Standard deviations within classes are high, as was
the case when actual traffic counts were evaluated.
Table 6. Mean of modelled traffic counts for classes
DRA
Class
Mean
Standard
Deviation
DRA Subclass
Mean
Standard
Deviation
Local
4,378
5,105
Local
4,033
5,061
Collector
5,377
4,412
Collector minor
3,899
3,200
Collector major
6,223
4,775
Arterial minor
6,853
5,298
Arterial major
12,314
8,942
Highway minor
9,762
8,194
Highway major
17,666
12,134
Freeway
39,431
6,358
Arterial
10,943
Highway
Freeway
8,463
15,478
12,225
13,546
10,612
70000
EMME/2 Predicted Traffic Count - GVRD
60000
10308
10305
50000
40000
30000
20000
10000
10292
10282
10280
10277
10266
10229
10226
10227
10223
10209
10207
10195
10192
10188
10189
10131
10132
10121
10119
10115
10110
10105
10099
10100
10087
10075
10053
10041
10034
10019
10004
9984
9977
9975
9964
9952
9926
9909
9906
9843
9828
9824
9821
9796
9792
9791
9782
9762
9743
9740
9739
9718
9716
9697
9691
9686
9683
9654
9631
9630
9613
9593
9592
9591
9582
9581
9572
9560
9551
9497
9493
9482
9479
9466
9452
9448
9446
9414
9405
9403
9402
9390
9366
9365
9352
9337
9334
9316
9287
9271
9258
9259
9253
9232
9226
9215
9214
9213
9204
9202
9193
9183
9179
9175
9168
9166
9164
9150
9115
9111
9086
9062
9053
9045
9044
9043
9019
9016
8998
8993
8982
8972
8968
8943
8938
8897
8896
8890
8885
8878
8879
8877
8876
8872
8860
8843
8839
8840
8838
8835
8832
8801
8771
8769
8768
8754
8755
8753
8745
8740
8730
8720
8715
8703
8679
8671
8658
8659
8645
8637
8638
8632
8628
8618
8613
8609
8606
8604
8599
8600
8593
8587
8578
8565
10228
10208
10181
10138
10313
10312
10311
10310
10309
10307
10306
10302
10301
10300
10298
10299
10297
10296
10295
10291
10290
10284
10283
10281
10279
10278
10275
10270
10271
10272
10265
10267
10268
10263
10262
10264
10260
10261
10259
10258
10249
10247
10245
10246
10244
10239
10238
10237
10234
10235
10230
10222
10224
10225
10217
10218
10216
10214
10215
10210
10200
10199
10201
10203
10204
10205
10206
10196
10197
10198
10193
10191
10190
10183
10184
10185
10186
10178
10179
10180
10170
10171
10173
10172
10175
10176
10174
10177
10162
10163
10164
10165
10166
10167
10169
10168
10158
10159
10160
10161
10156
10157
10149
10150
10151
10140
10141
10143
10145
10144
10134
10135
10136
10137
10139
10133
10304
10303
10289
10288
10285
10286
10287
10293
10294
10276
10269
10257
10248
10213
10212
10063
9969
9915
9914
9872
9867
9855
9842
9831
9832
9819
9807
9793
9781
9764
9710
9709
9707
9705
9706
9698
9688
9687
9679
9680
9655
9650
9628
9627
9620
9610
9598
9594
9578
9577
9567
9566
9557
9555
9533
9531
9527
9526
9520
9513
9501
9502
9503
9504
9494
9487
9473
9464
9458
9449
9447
9445
9422
9421
9407
9404
9397
9396
9393
9388
9389
9384
9367
9370
9368
9369
9358
9357
9343
9339
9338
9332
9330
9324
9318
9315
9314
9311
9301
9284
9282
9281
9276
9266
9265
9255
9252
9245
9244
9246
9237
9236
9230
9227
9228
9223
9221
0
N=
3161
3346
3448
Local
Collector
Arterial
242
116
Highway Freeway
Figure 4. GVRD EMME/2 traffic volumes by DRA classifications.
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
7
EMME/2 Predicted Traffic Count - GVRD
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
70000
60000
50000
40000
30000
20000
10000
10308
10305
10292
10282
10280
10293
10266
10229
10226
10227
10223
10207
10195
10192
10189
10188
10131
10132
10105
10099
10100
10087
10075
10034
10041
10019
10004
9984
9975
9964
9952
9909
9906
9843
9828
9824
9796
9792
9791
9782
9740
9739
9716
9697
9691
9686
9613
9593
9592
9591
9582
9581
9572
9560
9497
9493
9479
9466
9452
9446
9414
9405
9402
9403
9352
9287
9271
9226
9214
9213
9193
9183
9179
9175
9164
9062
9043
9019
9016
8993
8972
8968
8938
8897
8896
8890
8885
8879
8877
8878
8872
8843
8839
8840
8838
8835
8801
8771
8769
8754
8755
8745
8720
8715
8703
8679
8671
8658
8645
8638
8637
8632
8628
8618
8613
8606
8604
8505
8487
8474
8445
8444
8438
8437
8427
8422
8421
8415
8412
8377
8371
8348
8343
8342
8336
8327
8293
8285
8284
8283
8280
8279
8246
8241
8242
8226
8225
8224
8212
8197
8186
10017
9718
9615
9566
9369
9367
9370
9365
9368
9366
9338
9337
9276
9123
9074
9012
8746
8721
8700
8701
8683
8684
8614
8577
8576
8535
8536
8530
8531
8506
8441
8436
10228
10146
10148
10063
9954
9915
9914
9872
9867
9855
9832
9831
9807
9793
9781
9764
9710
9709
9707
9705
9706
9698
9688
9679
9680
9655
9650
9631
9628
9627
9610
9598
9594
9578
9577
9567
9557
9555
9551
9533
9531
9527
9526
9520
9513
9502
9501
9503
9504
9494
9487
10159
10115
10053
10054
9977
9951
9946
9917
9908
9870
9853
9842
9809
9810
9808
9801
9800
9795
9794
9787
9779
9778
9768
9763
9762
9760
9761
9753
9752
9751
9743
9736
9720
9719
9713
9683
9675
9671
9665
9663
9664
9654
10313
10312
10311
10310
10309
10307
10306
10302
10301
10300
10299
10298
10297
10296
10295
10291
10290
10284
10283
10281
10279
10277
10278
10275
10270
10271
10272
10265
10267
10268
10264
10261
10263
10262
10259
10260
10258
10248
10249
10247
10245
10246
10244
10243
10239
10238
10237
10234
10235
10230
10222
10224
10225
10217
10218
10215
10214
10216
10212
10304
10303
10285
10287
10286
10289
10288
10211
10202
0
N=
2421
1575
2249
1134
2692
44
192
6
ay
)
w
aj
ee
m
(
Fr
ay
)
in
hw
ig
(m
H
ay
hw aj)
ig
H
m
l(
ria
)
te
in
Ar
m
l(
)
ria
aj
te
m
Ar or (
)
ct
in
le
ol
(m
C
or
ct
le
ol
C
l
ca
Lo
Figure 5. GVRD EMME/2 traffic volumes by DRA sub-classifications.
Assuming that the sample of actual traffic counts is representative of true traffic volume, it is
possible to conclude that modelled traffic volume does not adequately predict differences in
volume between local and collector roads, or among arterial roads, highways, and freeways
(Table 7). Using modelled traffic volumes to indicate the level of exposure to traffic related air
pollution therefore could create substantial exposure misclassification. Another possibility is that
the sample of actual traffic counts is biased and coincidentally favours the DRA road class
system.
Table 7. A comparison of actual traffic counts and modelled traffic volume for
DRA road classes
DRA Class
Actual Count mean
Modelled Count mean
Local
3,976
4,378
Collector
8,953
5,377
Arterial
18,457
10,943
Highway
27,961
15,478
Freeway
113,789
13,546
Spatial Exploration of Variation in Traffic Counts by DRA Road Class. The variability of traffic
counts within each road class may be due to systematic traffic volume variation depending on
location within the study region. For example, it might be that a ‘local’ road in a high population
urban area always has higher traffic volumes than a ‘local’ road in a lower population rural area.
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
8
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
We chose three regions within the study area, based on population, population density1 and
available sample size for actual traffic counts (Table 8) and focused on local, collector, and
arterial road classes since these classes showed high internal variability.
Table 8. Characteristics of Compared Regions
Region
Population
Vancouver
Richmond
Abbotsford
583,296
172,714
126,634
Area
(hectares)
11,467
12,869
35,918
Population Density
(persons/hectare)
51
14
4
Source: BC Stats Community Fact Sheets
http://www.bcstats.gov.bc.ca/data/dd/facsheet/facsheet.htm
If there was systematic variation in traffic volumes based on regional differences in population,
we would expect to see high traffic volumes in high population areas, moderate traffic volumes in
moderately populated areas, and low traffic volumes in low population areas. Conversely, it
might be expected that increasing density decreases the need to use a vehicle, and therefore traffic
volume might decrease as population density increases. In fact, box plots of the mean traffic
count for local, collector, and arterial roads (Figures 6, 7 and 8) in Abbotsford, Richmond, and
Vancouver, suggest there is no systematic regional pattern. For local roads, Figure 6 shows
similar mean traffic counts in Richmond and Abbotsford, areas with similar populations but
different population densities, indicating no difference based on population density. The lower
mean traffic count in Vancouver, however, indicates there might be a difference based on
population density, assuming higher density causes lower traffic volumes. For both collector and
arterial roads, Figures 8 and 9 show similar means for low density and high density areas (and
therefore no difference based on total populations), but a lower mean for the moderate density
area (and therefore no difference based on total population or on density). Based on this limited
analyses, there is no evidence of a systematic spatial pattern in traffic counts that would explain
variability within each road class.
1
We note that in both Richmond and Abbotsford, substantial agricultural lands are included in the total
area, therefore population density may be substantially higher in some parts of each area. For example, if
we assume that 50 percent of Richmond is agricultural, population density could be 28 persons/hectare in
residential areas. Available traffic counts are, however, generally located in the most densely populated
areas of each region, and so it is likely reasonable to make direct comparisons among these regions.
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
9
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
10000
Actual Traffic Count - Local Road
8000
6000
4000
2000
0
N=
16
16
6
Abbotsford
Richmond
Vancouver
Figure 6. Actual traffic volumes on local roads in
Abbotsford, Richmond, and Vancouver
40000
18
17
Actual Traffic Count - Collector
30000
10
20000
8
10000
0
N=
17
11
26
Abbotsford
Richmond
Vancouver
Figure 7. Actual traffic volumes on collector roads in
Abbotsford, Richmond, and Vancouver
70000
60000
Actual Traffic Count - Arterial
50000
40000
30000
20000
10000
0
N=
6
23
16
Abbotsford
Richmond
Vancouver
Figure 8. Actual traffic volumes on arterial roads in
Abbotsford, Richmond, and Vancouver
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
10
UVIC SSL W ORKING PAPER 05-014
MARCH 2005
Conclusion
Traffic volume on nearby roads has been used to indicate exposure to traffic-related air pollution.
When few or no actual counts are available for a study area, road class may be used as a surrogate
for traffic volume, assuming that higher capacity roads carry more traffic on average than do
lower capacity roads. Modelled traffic volumes also may be used when actual counts are not
available.
Analyses of road classification systems used for existing digital road data in British Columbia,
and of modelled traffic volumes in the Lower Mainland suggest that, for the purposes of
assigning levels of exposure to traffic-related air pollution, the road classes used in the provincial
Digital Road Atlas provide a somewhat better differentiation in traffic volumes, than road
subclasses. Road classes used in DMTI digital road data did not sufficiently differentiate traffic
volume among classes. Modelled traffic volumes did not compare well to actual traffic counts
and when associated with road classes from the DRA digital roads, were not well differentiated
among classes. For the purposes of assigning exposure levels then, we conclude that at this point
in time, based on the sample of actual counts, using the DRA digital road class provides the best
chance for minimizing exposure misclassification, although more analysis of within class
variation is required.
We found no apparent spatial pattern that would explain the within-class variation and betweenclass overlap in traffic volumes, although our analysis was limited by the small sample size for
actual traffic counts for each region compared, and to the examination of the effects of total
population and population density on traffic volume. Undoubtedly, there are more factors, both
spatial and aspatial that may contribute to variation in traffic volume. Future research in this area
should attempt to identify additional factors that influence traffic volume, and use a larger sample
of actual traffic counts, optimized for each region included. If successful, these efforts could
produce a model suitable for predicting traffic volumes in regions outside of the Lowe Mainland,
based on the DRA digital road classes.
©SPATIAL SCIENCES LABORATORIES OCCASIONAL PAPERS SERIES 2005
11