The final step is to estimate AADT for the road section in question, where SPTC are available.
The degree that a road section belongs to each group, which is found in the output of the neural network, is used to calculate AADT.
For example, if the degree of belonging to road group 1 and (1 or 2) are
m(1)= 0.4 and m(1, 2) = 0.6, respectively, the final weights adopted for the
estimation are calculated as w(1)= 0.4 + 0.6/2 = 0.7 and w(2) = 0.6/2 = 0.3.
Therefore, the AADT estimate for a given SPTC is calculated by:
AADT= w(1) · SPTC · fij1+ w(2) · SPTC · fij1 (6.7)
where fij1and fij2are found in Equation 6.3.
In the case of d-days SPTCs, the estimation of AADT is repeated d times using the 24hr volume data. The final AADT estimate is the average of these AADT estimates. For example, for a 3-days (72hr) SPTC the final AADT estimate is:
AADTFinal = AADT1+ AADT2+ AADT3
3 (6.8)
where each AADT is the AADT estimated using the 24hr volume data for each of the d monitoring days.
Part III
Chapter 7
Case Study
The approach presented in Chapter 6 has been implemented in a real
world situation, testing its validity using traffic data. Different combination
of SPCTs were tested, evaluating the accuracy of AADT estimates obtained in each condition.
Given the road network and the AVC sites, the Fuzzy C-means algorithm was used to create road groups. SPTCs were extracted from the AVC sites, and AADTs were estimated from the SPTCs, based on the assignment to road groups given by the Artificial Neural Network. The estimated AADT value were compared with the actual AADTs of the AVC sites.
Results have been interpreted considering measures of uncertainty (dis- cord and non-specificity) and compared with those obtained by two ap- proaches proposed in previous studies.
7.1
Data Source
The case study traffic data were obtained from the SITRA (TRAnsporta- tion Information System or ”Sistema Informativo TRAsporti”) monitoring database of the Province of Venice.
Since 2000 the local Transportation Office is responsible of the monitor- ing activities taken on the rural road network of the Province of Venice. The Department of Structural and Transportation Engineering of the Univer- sity of Padova (Della Lucia 2000) collaborated to the design of the original program and is still involved in data management and validation processes. The design of the system was inspired by the procedure proposed by Traffic Monitoring Guide. Automatic Vehicle Classifiers (AVCs) were in-
stalled in a small number of road sections to permanently collect traffic
data (PTCs). Road groups were identified based on similarity of seasonal adjustment factors calculated from PTCs. SPTCs (48hr to 2 weeks) and STCs (2 to 12 weeks) were periodically taken in the other road sections, fol- lowing a detailed monitoring program. These road sections were assigned
to road groups and AADT could be estimated adjusting STPCs with the corresponding seasonal adjustment factors.
Variations of traffic patterns in different period of the year are accounted
using 18 seasonal adjustment factors. These seasonal adjustment factors are calculated based on the combinations of:
• 3 day-types (Weekdays, Saturdays, Sundays);
• 6 two-month periods (January-February, March-April, May-June, July- August, September-October, November-December).
Monitoring road sections are generally equipped dual loop systems. Since the network links are mainly two-lane roads, two AVCs were installed in each site, one for each direction of traffic flow. AVCs collect hourly traffic
volumes for different vehicle and speed classes (Tables 7.1,7.2). Hourly data
are aggregated to obtain some relevant traffic parameters for each road section, including:
• Annual Average Daily Traffic (AADT); • Annual Average Daytime Traffic; • Seasonal Adjustment Factors; • 30th highest annual hourly volume; • Peak hour factors;
• Traffic growth trends.
Table 7.1:Length Classes Adopted by SITRA Monitoring Program Class Length [m] Level of Aggregation
0 1 2
I L< 5.0 LU01 LU11 LU21
II 5.0 < L < 7.5 LU02 LU12 LU22
III 7.5 < L < 10.0 LU03 LU13 LU23 IV 10.0 < L < 12.5 LU04 V 12.5 < L < 16.5 LU05 LU14 VI 16.5 < L < 18.0 LU06 VII L≥ 18 LU07
Periodically the Transportation Office issues reports with traffic data
summaries for each road section, that can be used by pratictioners and other public agencies (e.g. Figure A.1 reported in the Appendix). More de- tailed data can be also obtained from the website of the monitoring program (http://trasporti.provincia.venezia.it/pianif_trasp/osservat.html).
Table 7.2:Speed Classes Adopted by SITRA Monitoring Program Class Speed [km/h] Level of Aggregation
0 1 I V< 30 LU01 LU11 II 30< V < 50 LU02 III 50< V < 70 LU03 LU12 IV 70< V < 90 LU04 V 90< V < 110 LU05 LU13 VI 110< V < 130 LU06 VII V≥ 130 LU07
Since 2000 the number of AVCs has increased, reaching a good coverage
of different types of road in the network. In 2012 the number of working
AVC sites is 39, that is 78 AVCs which monitor directional traffic volumes (see Figure A.2 reported in the Appendix).
The traffic data for the study are the volumes obtained for the year 2005 at 42 AVCs (see Figure A.3 reported in the Appendix). The remaining
AVCs have been excluded from the analysis since they were affected by
considerable amounts of missing data in some periods of the year due to vandalisms and equipment failures.
Some assumptions have been made in the analysis:
• Data from monitored road sections have been analysed separately for each direction, based on the findings of Tsapakis et al. 2011;
• Estimation of AADT has been done for passenger vehicles only. Pas- sengers vehicles data were divided by truck vehicles data, with refer- ence to a 5 m-length threshold. This choice was made following the indications given by the FHWA concerning the specificities of truck traffic pattens, as reported in section 4.5.