• About this site|
• Site index|
• Contact us|
• Jobs

Monitoring and Data Management Protocol: Environmental Indicators for Transport

Acknowledgements

1 Introduction

2 Why Monitor?

3 What to Monitor

4 Vehicle Fleet Composition Indicator

5 Travel Activity Indicator

6 Mode of Journey to Work Indicator

7 Travel Time Indicator

8 Glossary

Appendix A: Sampling Framework

Appendix B: Estimated Precision In Estimating Average AADT

Appendix C: Identifying the Road Use Category (or Traffic Pattern Control Group)

Appendix D: Week Multipliers and Relative Error for Each Road Use Category

References

5 Travel Activity Indicator

The environmental pressure of road transport is directly related to the amount of vehicle travel undertaken. This is reported as the annual vehicle kilometres of travel (VKT) by road vehicles. Traditionally, estimates of annual travel have been derived from the volume of fuel sold using empirical relationships between fuel use and distance travelled (litres per 100km). While this approach has been used for some considerable time, there were problems regarding the average distance fuel relationships and the volume of diesel fuel that was used for other purposes, e.g. in boats and machinery.

The current preference is for direct measurement of VKT. The annual vehicle kilometres of travel can be determined in two ways:

Road Based VKT - counting the number of vehicles travelling over each kilometre of the road network (VKT_R)

Vehicle Based VKT - recording the kilometres travelled by each vehicle (VKT_V)

The first, VKT_R, may be measured and reported at the local level by the Road Controlling Authority (RCA) by counting the traffic on the roads in the local area. The latter (VKT_V) is derived from data held at the Transport Registry Centre of the Land Transport Safety Authority. Because it is not possible to identify exactly where the vehicles undertook the travel, VKT_V. can only be reliably reported at a national level. It is, however, possible to obtain estimates at a regional level based on the assumption that the vehicles undertook their travel within the region where they are domiciled. An indication of where a vehicle is domiciled is given by either the vehicle ownership records or the location of the vehicle testing facility that issued the most recent warrant (or certificate) of fitness.

However, one advantage of VKT_V is that the measure can be disaggregated by vehicle type which if done provides analysts with a composite indicator. The composite indicator will be more powerful than the separate indicators (Travel Activity and Vehicle Fleet Composition).

5.1 Road-based vehicle kilometres of travel (VKT_R)

Road-based vehicle kilometres of travel (VKT_R) is monitored at the Territorial Local Authority level and the results aggregated to give regional and national level estimates.

5.1.1 How to monitor

To measure the annual VKT on a particular road it is necessary to count every vehicle that travels on the road throughout the year and multiply the total vehicle count by the length of the road section under consideration. This process would need to be done for every road administered by a particular Road Controlling Authority.

It is impractical to count the traffic on every road 24 hours a day, 365 days of the year. However, it is possible to obtain an estimate of VKT_R with known levels of precision by adopting an appropriate sampling strategy. In essence, the process involves undertaking sufficient traffic counts to reliably estimate the Annual Average Daily Traffic (AADT) on each road in a randomly selected sample of roads. This average traffic load is then applied to the length of road in the population.

5.1.2 Sample design

Determining the optimum sample of roads that must be monitored in order to obtain a reliable estimate of the VKT_R that is undertaken in a particular area is a complex exercise. The volume of traffic carried by roads varies substantially throughout New Zealand, and what may be a suitable sample for estimating the VKT_R of roads controlled by Southland District Council for example is unlikely to generate a reliable estimate if the same sampling frame is applied in Auckland City.

It is therefore necessary to develop a specific sampling frame for each of New Zealand's 73 Road Controlling Authorities. These sampling strategies must take account of the variation in traffic flows that occur across an area, which in turn requires some prior knowledge of the traffic flows, the very thing that the strategy seeks to determine. However, as part of a recent project undertaken by Transfund New Zealand (Opus International Consultants 2002) sampling strategies have been developed for every Road Controlling Authority in New Zealand (excluding Chatham Island District Council) and these are available for use in monitoring this indicator.

To optimise the sampling strategy, the roads in any RCA are divided into five nominal traffic flow bands. These bands are based on the expected level of traffic flow, the variation in expected traffic flows across the band, and the total contribution of the band to the overall estimates of VKT_R. The traffic flow bands are defined by the percentage of the maximum traffic flow recorded in 1994 as shown in Table 4.

Table 4: Traffic flow bands

Although the base date of 1994 seems somewhat outdated, this was the year in which the most comprehensive study of national traffic levels was undertaken. As a result, the potential sampling errors could be minimised by standardising on this year and adjusting more recent data back to this date.

The sampling strategy is based on taking sufficient samples from each band in order to estimate the Average Annual Daily Traffic (AADT) for all roads in the band. The boundaries for each band have been determined individually for each RCA as have the number of samples to be taken from each band. The sample size required to estimate the VKT_R in each RCA to within approximately ±10% are given in Appendix A, together with the traffic flows that define the bands.

The following example (Table 5) shows that for Kaikoura District Council traffic volumes must be measured on a total sample of 19 roads:

• One traffic count is required on roads in Band 1 where the upper flow limit is 28 vehicles per day.
• One traffic count is required on roads in Band 2 where the traffic flows are expected to lie between 28 and 69 vehicles per day.
• Three traffic counts are required on roads in Band 3 where the traffic flows are expected to lie between 69 and 247 vehicles per day.
• Two traffic counts are required on roads in Band 4 where the traffic flows are expected to lie between 274 and 493 vehicles per day.
• Thirteen traffic counts are required on roads in Band 5 where the traffic flows are expected to lie between 493 and the maximum flow of 3090 vehicles per day.

Table 5: Example of sampling framework

View example of sampling framework (large table)

This sample framework provides estimates of AADT to within the confidence limits set out in Appendix B depending on the type of traffic counting used. Should a particular RCA require a more rigorous traffic monitoring framework for its own purposes, a sampling framework based on ± 5% has been developed as part of the traffic information database project undertaken by Transfund New Zealand. That project provides a CD ROM which contains the individual sampling frameworks for each RCA based on starting points of both ± 10% and ± 5% at the 95thpercent confidence limits. This information is discussed in more detail in section 5.1.3 below. However, a general rule is that to halve the confidence interval from ± 10% to ± 5% requires a four-fold increase in the number of monitoring sites (e.g. for Kaikoura the sample size would increase from 19 to 76).

5.1.3 Selecting monitoring sites

Although the majority of road controlling authorities have traffic monitoring programmes in place, these programmes typically target higher volume roads in this area. Consequently, reliance on these existing programmes to supply the necessary data will result in biased estimates.

In order to make a random selection of monitoring sites, it is necessary to have a database containing all roads in the RCA together with an estimate of the expected traffic volume on each. As part of the traffic information database project commissioned by Transfund New Zealand such a database was created for all RCAs and areas.

The traffic link database was compiled by linking the year 2000 road asset maintenance management (RAMM) system data from each RCA to the New Zealand road centreline database held by Critchlow & Associates. The traffic link database represents the road network as a series of traffic links - road sections having essential similar traffic flows. Each traffic link is described by a series of variables including:

TRAFF_ID a unique link identifier

START_ID and

END_ID a series of unique nodes that connect links

TLA the territorial local authority in which the road is located

RAMM_ID the RAMM road section number in the TLA's RAMM database

START_D and

END_D which record the RAMM start and end distance along the RAMM_ID

CAL_NAME the legal road name from the Critchlow & Associates Limited road centreline database

LENGTH the link length in metres

RUC the road use category which describes the seasonal variation in traffic flows

AADT_94 the estimated annual average daily traffic in 1994

This database can be obtained by contacting the:

Planning and Evaluation Group
Transfund New Zealand
PO Box 2331
Wellington.

The following procedure uses the above database to select an essentially random sample of sites where traffic levels should be monitored and avoids the bias inherent in existing programmes. The procedure is as follows:

Step 1 Obtain a copy of Transfund New Zealand's Traffic Information Database CD Rom.

Step 2 The Excel spreadsheet total samples.xls (on the CD) provides the number and type of samples required in each of five flow bands, together with the flow band limits to determine the average AADT for all roads administered by the RCA. There is one page for each RCA providing sample sizes for both a sample error of ± 10% and ± 5%. See example in Figure 1. It should be noted that the spreadsheet shown in Figure 1 suggests how the required sample in each band should be distributed over the various road use categories. This is a suggestion rather than a necessary feature of sample selection.

Step 3 The spreadsheet link list by TLAs.xls provides a list of all road links in the TLA together with details of road name, length, road use category and the base year AADT (1994). See Figure 2.

Step 4 The list of roads in Figure 2 can be used to select the monitoring sample:

• Sort the data by road name and remove all ACCESSWAY & SERVICE LANE records.
• Allocate each road section to the appropriate flow band based on AADT (using a suitable Excel macro).
• Allocate a random number to each record (using the excel f functions random).
• Sort the resulting file using the following multiple criteria (using the excel Data Sort):
  • Flow Band >>Descending
  • Road Use Category >>Descending
  • Random Number >> Ascending
• From the file, select the required sample by taking the records from the file in the order these appear, as shown in Figure 3 on the following page.

The example in Figure 3 shows the selection of 13 road sections in Flow Band 5. Of these it was recommended in the sampling framework that 8 be selected from Road Use Category 1b, two sites be selected from Road Use Category 2 and 1 each be selected from Road Use Categories 3, 5, and 9.

While the above procedure provides a means of selecting a random sample, it may also be modified in order to maximise the value of any existing traffic monitoring programme. It is acceptable to substitute an existing traffic count site in the place of a sample site provided that the base year (1994) AADT recorded at the existing count site is within ± 10% of the selected site.

In all cases, the sites should be visited to ensure that it is actually possible to undertake surveys at some point on the selected length of road. The preferred site should then be recorded and, if necessary, the actual location of the survey site photographed to ensure that future surveys are always undertaken in the same place.

5.1.4 Collecting the data

Having selected the monitoring sites, the next step is to measure the annual traffic flows at these sites. However, measuring traffic volumes continually throughout the year requires sophisticated counting equipment and data management systems. Such systems are expensive and costly to operate. The alternative is to undertake shorter duration traffic counts in order to estimate the average annual daily traffic at a particular location and multiply these values by 365 days to obtain the total traffic using the road in a year.

However, the daily traffic on a road varies systematically, depending on the day of the week. For example, a road servicing a predominantly industrial area will experience lower traffic volumes on Saturdays and Sundays. Similarly, weekly traffic volumes will vary systematically throughout the year, depending on the type and function of a road. For example, a road that services a seaside village is more likely to experience higher traffic volumes over the summer holiday months. If the estimate of AADT for this road is based on a single week's traffic count, undertaken during the summer holiday period, the resulting estimate of AADT will be considerably higher than an estimate of AADT based on a week-long traffic count undertaken during the cooler winter months.

One way of overcoming this problem is to undertake sufficient traffic counts randomly allocated throughout the year to establish a reliable estimate of the AADT. Typically, the variation in weekly traffic volumes on a particular road is considerable with 95% confidence intervals generally lying in the range of ± 10% to ± 30%. To be 95% confident that the true AADT lies within ± 10% of that estimated from a series of the traffic counts, four- to eight-week-long traffic counts would be required for each road, each year. The costs associated with such a programme would be substantial.

However, research by Transit New Zealand (Transit New Zealand 1994) and, more recently, by Transfund New Zealand (Transfund New Zealand 2001), has specifically investigated patterns of seasonal variation in traffic flows for different road types. By classifying roads in terms of their expected seasonal variation, it is possible to significantly reduce the number of traffic counts that are required to obtain a reliable estimate of AADT.

In short, this process involves allocating roads to one of the nine road use categories described in Table 6, using the procedure outlined in Appendix C. The pattern of seasonal variation for each category has been determined from an analysis of continuous traffic count data collected at similar sites throughout New Zealand. Using this data, it has been possible to compile multipliers, or expansion factors, that may be applied to a week-long traffic count undertaken in a known week to provide an estimate of AADT and the relative error associated with that estimate. These factors are given in Appendix D. One set of factors is presented for total traffic flow, the second set is for the heavy vehicles only.

Table 6: Road use categories

5.1.5 Types of Traffic Counting

The actual business of counting the vehicles using a particular road can be undertaken in a number of different ways, each of which provides slightly different data, although it is possible using a series of assumptions to obtain some standard data. The most common methods and their respective outputs are discussed below.

Visual Traffic Counts

Carried out by suitably trained or experienced staff, these surveys record the number of vehicles of different types that pass the observer. Typically, such surveys are undertaken for short durations in daylight hours. The classification of vehicles is based typically on the body type and may be customised to the particular requirements of the survey.

Two common classification systems are those used by the Land Transport Safety Authority and that used to obtain data for project evaluations using the Transfund New Zealand classification system. Examples of these classification systems are given in Appendix C.

Axle Counts (Tube)

These are undertaken using automatic recording systems that count the number of axles that pass over a rubber tuber across the road. Typically the axle counters record the number of axles passing over the tube, divided by two to represent vehicles, i.e. the assumption is that all vehicles have two axles. Clearly this is not the case, and a correction needs to be applied. Converting the axle count into vehicles requires determining the vehicle factor, which is defined as:

V_f = 2/A_f

where A_f = the axle factor and is the average number of axles per vehicle.

The axle factor must be determined for each site using a visual survey of 300 vehicles (Transit New Zealand 1994). The disadvantage is that the vehicle/axle factor recorded from the visual surveys is assumed to apply throughout the whole of the survey period and results in an average error of around 2-3%.

Vehicle classifiers (tubes)

This method uses two pneumatic tubes to record the number and spacing sequence of axle passes, and based on this data seeks to identify the type of vehicles that passed over the site. Individual classifier manufacturers of the equipment have different default categorisations, however, the most commonly used is the TNZ Class system which provides 13 classifications of vehicles and a 14th classification 'unknown'.

Although manufacturers' claims suggest the classifiers may accurately record vehicle types of high volumes of bi-directional traffic, it is important to ensure the number of 'unknown' vehicles remains small (<2%). If this is not the case one classifier may be needed for each lane of traffic.

Vehicle classifiers (induction loops)

This indirect method of recording traffic relies in the change of electromagnetic induction as a vehicle passes over detection equipment embedded within the pavement. This process measures, and consequently classifies, vehicles on the basis of body length. The typical configuration records four vehicle length are:

• 0-5.5 m small
• 5.5-11 m medium
• 11-17 m long
• >17 very long.

Classification systems

Unfortunately there is only limited consistency across the different vehicle classification systems used in New Zealand as shown in Table 7. Given that each TLA will have different equipment and local practices, it is recommended that the data be collected at the level of detail applicable to the equipment and systems used by the TLA. However, it should be reported at a common level based on the split between light and heavy vehicles.

Where data is collected by inductive loop classifiers which report four length-based classifications, it has been found that approximately 50% of vehicles in the 5.5-11 metre length bins are light vehicles, and the remaining 50% are heavy vehicles.

Table 7: Vehicle classification systems commonly used in New Zealand

View classification systems commonly used in New Zealand (large table)

5.1.6 Monitoring process

The monitoring process is essentially very simple. Once the sample sites have been selected, each is surveyed for a seven-day period. Ideally this period should be a week beginning on Monday at 0:00 (midnight Sunday) so that the appropriate week factor is used to adjust the counted traffic volume to the estimated AADT. Other start times are acceptable however specific week factor will need to be calculated as the day weighted mean of the weeks that the traffic count spans. For example, if a traffic count begins on Wednesday of Week 6, which has a factor of 1.0206, and ends at midnight on the following Tuesday in Week 7, for which the factor is 0.9737, then the appropriate factor for that traffic count is:

5/7 (1.0206) + 2/7 (0.9737) = 0.7920 +0.2782 = 1.002

Although vehicle classification surveys using either loop or tube based vehicle classifiers are preferred, the individual traffic surveys may be undertaken using any of the methods above. However, if axle counts are used a visual survey of at least 300 vehicles should be undertaken to determine the vehicle-to-axle ratio and the approximate percentage of heavy vehicles.

The data from the individual classifications counts should be aggregated to give the total volume of light and heavy vehicles at the site. This data is then input into the 'AADT Calculation Sheet' presented below.

ADDT Calculation sheet (large table)

Using the AADT Calculation Sheet, the estimated AADT of each sample length is determined for each site. Multiplying the recorded traffic flow by the multiplier from Appendix D does this. It is, however, important to note that separate multipliers are given in Appendix D.

(i) Total traffic and

(i) Heavy vehicles.

By implication, the volume of light vehicles is the total volume minus the volume of heavy vehicles.

Once the estimated AADT for each of the sample sites has been determined, it is necessary to compute the VKT for each of the sampled links and to sum this over all samples in a particular flow band and the total length of the sample links in each band. This is done using the Sample VKT Calculation Sheet (over page). One sheet should be used for each band.

View sample VKT Calculation sheet (large table)

The total sample VKT and total road length in the sample is obtained by summing over all sample sites in the band, while mean AADT for total and heavy vehicles is also computed for the sample sites in each band.

The total road length in the sample together with the sample VKT are transferred to the Calculation of Annual VKT Sheet where the sample VKT in each band is scaled up to represent the total VKT for all roads in the sample band.

This scale factor for length is determined for each band by dividing the total road length in the sample of that band by the total road length in the band within the TLA. The latter can be easily obtained using the pivot table capability in Microsoft Excel to sum the total length of road in each band.

View calculation of Annual VKT sheet (large table)

5.1.7 State highways

The method described above provides an estimate of annual VKT for those roads administered by the TLA. It does not estimate the VKT undertaken on state highways within the TLA. The VKT on state highways must be computed and recorded separately. This avoids the possibility for double counting of this volume.

Transit New Zealand's traffic monitoring system (TMS) records traffic volumes on all sections of state highway. The individual traffic counts are automatically corrected for seasonal variation as each individual traffic count site is linked to a continuous automatic traffic count site. While TMS provides a breakdown of highways by regional government area it does not include a breakdown by TLA. However, it is a relatively simple matter to identify the break points for each section of state highway and obtain the link length. The traffic flows contained in Transit's TMS system are exported to populate Transit's RAMM database. Multiplying the RAMM road section lengths by the AADT provides the VKT on each road section. Using the TLA code in the RAMM carriageway table it is possible to calculate the state highway VKT in each TLA.

Contact Transit New Zealand to obtain this information.

5.1.8 Reporting

Table 8 sets out the data that should be reported on VKT annually for the TLA. This data should be reported as millions of vehicle kilometres travelled. The number of sites monitored by the TLA should also be recorded in a footnote to the data together with the nominal precision of the sampling, e.g. 95% confidence limit ±10% at TLA level.

Table 8: Reporting requirements for VKT_R

Once consistent monitoring has been undertaken for a number of years changes in VKT over time can be identified and trends determined.

5.1.9 Data management

The calculation sheets presented in the monitoring process contain all the necessary data to monitor and report VKT. These forms, or electronic copies of the forms, should be stored for future reference, and to create a suitable time series. This data and the associated summaries are all that need to be stored for reporting VKT.

One feature of the sampling framework is the assumption that the specified road use categories and base year traffic data (AADT 1994) are correct. It must, however, be recognised that some errors in this assignment are likely to have occurred. Over time, it is likely that each TLA will have reason to measure the traffic on road sections other than those required by the sampling framework. These measurements will allow the classification of links to be confirmed and adjusted if necessary. These adjustments should be recorded in the TLA's own copy of the link data set so that further updates of the main database can be undertaken in future years if necessary.

One important use for traffic count data is to populate the TLA's RAMM database. While many TLAs undertake traffic counts for this purpose, the counted traffic volumes are seldom adjusted to provide reliable estimates of AADT. Although it is beyond the scope of this monitoring it is strongly recommended that the TLAs use a system based on the AADT Calculation Sheet to record traffic counts and to ensure that the count data and appropriate multipliers are used.

It should also be noted that the Transfund New Zealand traffic information database project provides a mechanism by which single 24-hour traffic counts or even part day three-hour traffic counts may be factored to provide a reliable estimate of AADT. This process may be used in future years. However, for the immediate task it is necessary to confirm that the selected sample sites are allocated to the correct road use category. This can only be done with seven-day traffic surveys.

5.2 Vehicle based vehicle kilometres of travel (VKT_V)

Although it is possible to monitor the environmental pressure of transport activities by considering separately volume of travel and the characteristics of the fleet undertaking that transport activity, it is the combination of the two that is the most informative measure. Vehicle based measures of travel activity, VKT_V, provide this data. In addition to acting as a check on the volume of travel activity derived from road based measurements, VKT_V will assist in identifying whether increases in travel are associated with particular vehicle types that may have a greater or lesser impact on our environment. It is, however, extremely difficult to determine where this travel was undertaken except in the broadest sense.

5.2.1 How to monitor

Since 1995 the vehicle odometer reading has been recorded as part of the periodic warrant of fitness (or Certificate of Fitness) process, and is held in the Certificate of Fitness/Warrant of Fitness database. In 1997, the LTSA linked the database for certificates and warrants of fitness to the vehicle licence and registry databases. It was mandatory to provide evidence of a valid safety inspection prior to automating the process. It then became mandatory to have a valid certificate of fitness at the time of relicensing. Odometer data is now available based on the annual inspections for vehicles up to six years old and the six-month inspections for older vehicles. This data is available from the LTSA. Vehicle based VKT is established by determining the annual kilometres travelled by each individual vehicle and then summing over all vehicles of a particular type, as described in Section 4.

While the process appears relatively simple in principle, there are a number of confounding factors, including:

1. alignment of the reporting period with the inspection period
2. variable recording frequency depending on vehicle age
3. treatment of new vehicles, those which have been operating for less than 12 months and have therefore not been subject to inspection and
4. poor data quality, missing information, transposed figures and negative records when, for example, the odometer range is exceeded and the system begins from zero.

To overcome these issues it is first necessary to identify those vehicles in any particular group for which the data appears invalid. Then, using the remaining data estimate the mean travel undertaken by vehicles in that group. This value is then applied to the total number of vehicles in the group. Information on the number of vehicles in the 'poor-quality data' scenario can be advised.

This is the process adopted by the Research and Statistics Unit of the Land Transport Safety Authority. This unit is currently undertaking this exercise and plan to report National Estimates VKT_V on a regular basis.

Agencies wishing to report VKT_V should contact:

The Manager
Research and Statistics
Land Transport Safety Authority
PO Box 2840
Wellington

5.2.2 What to monitor

The aim of measuring VKT_V is to form a composite measure that combines VKT with fleet composition. Accordingly the vehicle types that should be monitored are categorised in the same way as the vehicle fleet indicator. These vehicle types are:

1. Two wheeled power vehicles
   moped VTYP = 01
   motorcycle VTYP = 11
2. Passenger cars and vans VTYP = 07
3. Light goods vehicles VTYP = 08 GVM <3.5 tonnes
4. Medium goods vehicles VTYP = 08 3.5 < GVM < 7.5 tonnes
5. Heavy goods vehicles VTYP = 08 GVM > 7.5 tonnes
6. Buses VTYP = 09
7. All others (except unpowered trailers)

GMV = gross vehicle mass.

Although it may well be possible to also break down the passenger car and van fleet on the basis of engine capacity and fuel type used, this is considered excessive detail for an indicator although such detail may be useful in policy specific analysis.

5.2.3 Reporting

As with the vehicle fleet indicator, this data can be reported nationally and regionally based on the assumption that the owner details and/or the warrant (certificate) of fitness testing station is an indicator of where a vehicle is domiciled. The annual VKT should be reported together with the Vehicle Fleet Indicator variables as shown in Table 9.

Table 9: Example of reporting vehicle kilometres travelled (vehicle based)

View example of reporting vehicle kilometres travelled (large table)

5.2.4 Data management

The most pressing issue with data management is ensuring that the odometer readings are valid without introducing bias. While it is reasonable to assume that data indicating vehicles have undertaken negative travel are invalid, there must be some suspicion regarding those that travel either very low or very high distances annually. The first step is to calculate the mean travel for each group. Ideally using data from at least three inspections, it is possible to obtain two travel intervals which will provide sufficient history to align the inspection period to the analysis period. However, where a vehicle is inspected only annually, it is not possible to obtain this data for all vehicles. To over come this it is recommended that in each vehicle type group three subgroups:

• 'new' vehicles, those for which there is no history
• 'relatively new' vehicles, those for which annual tests are required
• 'older' those which are tested at six-monthly intervals.

For each subgroup calculate the subgroup mean annual travel and standard deviation. Then identify all vehicles that are reported to have travelled distances more or less than three standard deviations either side of the mean for that group and discard from the sample. Likewise it is not possible for a vehicle to travel exactly zero km between inspection, so these vehicles should also be removed from the sample. Once these are discarded from the sample, recalculate the mean and apply this mean annual travel to all registered vehicles in the group (i.e. including those not included when calculating the mean for the group). For the 'new' vehicles use the annual mean for the 'relatively new' vehicles.